Help talk:Citation Style 1

Citation templates ... in conception ... and in reality

Cite book Harv warning

Initial conversation

Cite book has started giving a harv error message even when ref=harv has not been added. E.g.

Ratcliffe, Derek, ed. (1977). A Nature Conservation Review. Vol. 2. Cambridge, UK: Cambridge University Press. ISBN 978-0-521-21403-2.

It only did this before with the citation template. Many articles use this template without ref=harv, and they are all suddenly giving error messages, which is ugly and distracting. If it carries on I will have to turn off the harv warnings, but then I will not be warned when I am using ref=harv and I make a mistake. What is the position? Dudley Miles (talk) 21:54, 18 April 2020 (UTC)

It's not giving an error message for me? Have you got some script running?Nigel Ish (talk) 22:02, 18 April 2020 (UTC)

Yes, he has, the usual one. In User:Dudley Miles/common.js, there is a line calling User:Ucucha/HarvErrors.js. That script complains about every citation or cite template that uses ref=harv and is not referred to by a harv template, even when there is no actual error in the citation or cite template. It is the script that is the problem, not the cite book template. Remove it from your common.js and the error messages will go away. Or I have a modified copy in User:David Eppstein/HarvErrors.js that I use, to detect other errors but ignore this non-error. —David Eppstein (talk) 22:09, 18 April 2020 (UTC)

(edit conflict)

User:Ucucha/HarvErrors. See the documentation there for how to turn off the warning messages that that script emits for templates that do not have anything pointing at them.

See also Category:Harv and Sfn template errors

—Trappist the monk (talk) 22:14, 18 April 2020 (UTC)

Change User:Ucucha/HarvErrors.js for User:Svick/HarvErrors.js and that will take care of spurious warnings, but leave error-detection in place. Headbomb {t · c · p · b} 22:30, 18 April 2020 (UTC)

I've come here to ask about the same thing. It's suddenly doing what {{Citation}} does. Someone must have made a change. SarahSV ^(talk) 23:27, 18 April 2020 (UTC)

Yep: Help talk:Citation Style 1 § module suite update 18–19 April 2020 and in particular: Help talk:Citation Style 1 § make ref=harv the default for CS1.

—Trappist the monk (talk) 23:54, 18 April 2020 (UTC)

Please undo the edit(s) that caused the change. SarahSV ^(talk) 00:11, 19 April 2020 (UTC)

For people who use this template in Selected works and Further reading, it's going to cause a mass of error messages. SarahSV ^(talk) 00:21, 19 April 2020 (UTC)

@SlimVirgin: it is going to cause a script that you personally installed for your own use to incorrectly display error messages for things that are not errors. This incorrect behavior of the script is old and not the fault of the citation templates. The citation templates should not be forced to work around your idiosyncratic choice of script installation. —David Eppstein (talk) 00:26, 19 April 2020 (UTC)

(edit conflict)

My position in the discussions preceding this particular change did not prevail. You are using User:Ucucha/HarvErrors so you can turn off the warnings. Add this line to User:SlimVirgin/common.js:

window.checkLinksToCitations = false;

—Trappist the monk (talk) 00:31, 19 April 2020 (UTC)

Trappist the monk, thank you. SarahSV ^(talk) 00:34, 19 April 2020 (UTC)

Trappist the monk, I'm no longer getting the warning messages if a source in Works cited isn't used, what Gerda below calls the wanted error messages. This change is not good. Things were working well. SarahSV ^(talk) 02:47, 21 April 2020 (UTC)

@SlimVirgin: then use User:Ucucha/HarvErrors.js and you should have those. Headbomb {t · c · p · b} 02:59, 21 April 2020 (UTC)

Let me understand: wouldn't that also "kill" the wanted error messages, where I misspelled a name, for example? - With the script in place, the examples for {{cite book}} get messages. I just watched a user adding |ref=none to further readings. Not what I want to do for thousands of articles, and making every older version still look horrible. --Gerda Arendt (talk) 06:41, 19 April 2020 (UTC)

With the version of the script I'm using (or with the modification described above), I still get error notifications for actual errors: when a harv ref refers to a link that is not there. That covers both the case of a misspelling in the harv link and a misspelling in the reference itself. I merely don't get errors for articles that have ref=harv citations (usually because they use citation style 2 for which this has long been the default) but don't point to them using harv templates, something that is not actually an error. If I saw someone mass-adding ref=none on my watchlist I would revert them and tell them to fix their user script. —David Eppstein (talk) 06:50, 19 April 2020 (UTC)

(ec, a bit of patience please with old users who never heard of citation style 2, - not an answer therefore) I wrote many book definitions without ref=harv precisely if I did NOT want to cite them. I understand the default was changed, and I use the wrong script. Could a bot perhaps kindly go over Further reading sections and External links and add |ref=none because I don't want to annoy readers who use the "wrong script" which worked well for me? ... or change the script? ... or at least tell users "You use the wrong script"? ... or perhaps replace it? --Gerda Arendt (talk) 07:06, 19 April 2020 (UTC)

@Gerda Arendt: might be proposed at WP:BOTREQ – I already posted a message there, see WP:BOTREQ#Cleanup of cite templates after "ref=harv" became default and/or update to HarvErrors.js (which means: suggesting to keep the discussion here for the time being, until a clear botreq can be formulated). --Francis Schonken (talk) 07:39, 19 April 2020 (UTC)

Thank you for trying but the reply above looks as if such a bot would be reverted ;) --Gerda Arendt (talk) 07:47, 19 April 2020 (UTC)

Sorry I'm late, I posted the following at Template talk:Citation, and I appreciate much of this has been said and argued over already:

"This week, for the first time ever that I know of, ordinary {{cite web| ... and {{cite new| ... and probably other such templates are suddenly displaying warnings like

"Harv warning: There is no link pointing to this citation. The anchor is named CITEREFLaSala2016."

This type of message used to occur, helpfully, only when there was an explicit Harvard parameter indicating that links were expected: ... |ref=harv }]

That was useful as one normally only put the harv indicator in when the plan was to insert inline citations in the text, and missing one was obviously a problem.

Now, however, the error occurs WITHOUT the ref=harv parameter! This makes no sense: it causes an error if any citation is listed in a bibliography or further reading, and worse, if citations are listed inside an inline ref tag --- it seems to happen from the third citation onwards --- there is a fine specimen at K. Pattabhi Jois, and there is no visible means of suppressing the errors.

This makes checking for ACTUAL errors much more difficult, as false positive "errors" are displayed all over. Would be glad if the status quo could be resumed, please." Chiswick Chap (talk) 12:18, 21 April 2020 (UTC)

@Chiswick Chap:, there is no error caused by this. As pointed out by the big notice at the top, what you see here is User:Ucucha/HarvErrors.js complaining about CS1 templates and throwing "warnings". If you don't like those, you are welcomed to use User:Svick/HarvErrors.js, or request a new script made at WP:SCRIPTREQ that behaves like you want. Headbomb {t · c · p · b} 12:46, 21 April 2020 (UTC)

With respect, I beg to differ. The "errors" in the old dispensation were immensely useful in locating broken links. The new "errors" are a b. nuisance. I'm sure I'm not the only editor "old and wise enough to know the difference". I don't with respect want to switch all of them off (or on): I want the old arrangement which worked correctly. Chiswick Chap (talk) 12:56, 21 April 2020 (UTC)

The "old" arrangement didn't "work correctly". It worked "correctly" on a handful of articles that were specifically written around this script. I know, because I used the script before the update, and the "warnings" it threw were plentiful and irrelevant 9 times out of 10. Now every citation is treated the same, which gives greater flexibility in how people can write articles without being unnaturally forced into CS1 or CS2. Make a WP:SCRIPTREQ, and someone can design a new script that behaves like you want. Headbomb {t · c · p · b} 13:48, 21 April 2020 (UTC)

In other words, dear Chiswick Chap and all other editors, you people are far too dim and ill-informed to understand and appreciate the FINER things of life, art, and engineering, and should HUMBLY and MEEKLY accept the PEARLS OF WISDOM so GENEROUSLY dispensed to you from above by THOSE WHO KNOW. The fact that you miserable serfs fellow-editors find it b. inconvenient is merely a passing problem in the ways of Wikipedia, a small ripple in space-time, a miniscule play in the eternal Lila of the illusions of the world, and you should quietly f. off and die accept that if you had greater intellect you would dimly glimpse the splendours of cascading CSS sheets, the marvels of Perl scripts, and the untrammelled bliss of pure machine code running unhindered on a whole farm of servers, subservient to the unified and resplendent will of the UNIVERSAL PROGRAMMER. Chiswick Chap (talk) 13:58, 21 April 2020 (UTC)

Or, you know, go to WP:SCRIPTREQ and ask for an updated script. Headbomb {t · c · p · b} 14:49, 21 April 2020 (UTC)

(edit conflict)

If by third citation onwards you are referring to this group of references, the first two of that group do not have contributor, author, or editor names. An anchor ID is the concatenation of CITEREF, up to four contributor/author/editor names and a date. When there are no names, no anchor ID (this is not new).

You might replace User:Ucucha/HarvErrors.js with User:Trappist the monk/HarvErrors.js. That script does not show the warning messages unless the article uses one of the {{harv}} or {{sfn}} families of templates.

—Trappist the monk (talk) 12:58, 21 April 2020 (UTC)

@Chiswick Chap: If you want to find broken links, without the warnings, then use User:Svick/HarvErrors.js. Headbomb {t · c · p · b} 13:15, 21 April 2020 (UTC)

Sorry to join in with this late. Looking at this from a different angle, automatically adding harv anchors to every cite template not immediately proceeded by a <ref> tag is adding unnecessary html to the page when parsed. Whilst this isn't a problem with only a few instances on the page, on certain pages with maybe 50 "Further reading" links, this will add a significant overhead to the html (which could be problematic on older mobiles). When the page is parsed to html, the harv anchors become html anchors with an id equal to the CITEREF anchor. In html, id's must be unique. When we had to add anchors manually, where we cite the same author with several publications in the same year we differentiate the anchors eg Smith 1990a; Smith 1990b; Smith 1990c; etc. Automatically adding the anchors, would in the above case give 3 anchors with an id of "Smith 1990", causing non-compliant html. This probably isn't a problem most of the time, but the worst case scenario would be the browser going into "quirks mode", stop rendering the page and then redraw the page once the html, css, javascript etc has finished downloading. Again, this is more likely to happen on mobiles. As 40 - 70% (dependent on who's figures you look at) of internet access is from mobiles, then how the site behaves on mobiles needs to be considered. Putting aside "errors" being shown up by scripts, we should be adding |ref=none to unused anchors to prevent html problems. --John B123 (talk) 16:05, 15 June 2020 (UTC)

automatically adding harv anchors to every cite template not immediately proceeded by a <ref> tag is adding unnecessary html to the page when parsed

Perhaps. But, it is very common for editors who use short-form citation templates to have a Bibliography section populated with cs1|2 templates, none of which are wrapped in <ref>...</ref> tags and many of which are the targets of the short-form citations. In fact, I suspect that the opposite is true; in most cases, cs1|2 templates wrapped in <ref>...</ref> tags don't need anchor IDs because generally, nothing links to them except the superscript added by cite.php during conversion from wiki-text to html.

where we cite the same author with several publications in the same year we differentiate the anchors eg Smith 1990a; Smith 1990b; Smith 1990c; etc. Automatically adding the anchors, would in the above case give 3 anchors with an id of "Smith 1990"

Only true if the en.wiki editor has been lazy and failed to disambiguate |date= in the cs1|2 templates as should be done when citing three 1990 vintage Smith sources. The lazy en.wiki editor should have written {{cite ... |last=Smith |... |date=1990a |...}}, {{cite ... |last=Smith |... |date=1990b |...}}, {{cite ... |last=Smith |... |date=1990c |...}}. If the article is using short-form citation templates to link to the long-form cs1|2 citations, Module:Footnotes detects cs1|2 templates with identical anchor IDs and, when there is a short-form citation that uses that anchor ID, emits an error message. These error messages are hidden; see :Category:Harv and Sfn template errors § Displaying error messages to learn how they may be displayed for you.

I do not dispute your claim that rendering an anchor ID with every cs1|2 template adds a few bytes per citation to a rendered article. Where that is a concern, mostly I would think in very large articles with hundreds of cs1|2 templates, it is possible to turn-off the anchor ID for each cs1|2 template with |ref=none. There has been some discussion about creating a cs1|2 configuration template that would be read when the wiki-text is converted to html that might be used to disable all cs1|2 anchor IDs except those specifically created (|ref= has a value other than none). As far as I know, no one is clambering for such a template.

—Trappist the monk (talk) 17:21, 15 June 2020 (UTC)

In fact, I suspect that the opposite is true; in most cases, cs1|2 templates wrapped in <ref>...</ref> tags don't need anchor IDs because generally, nothing links to them - My point in mentioning the <ref>...</ref> in my previous post was exactly that, the issue generally occurs outside the references section: Bibliography, Further reading etc. (On a side note, where multiple cite templates are bundled within single <ref>...</ref>, anchors are added for the second and subsequent cites.)

Only true if the en.wiki editor has been lazy and failed to disambiguate |date= in the cs1|2 templates I'm not sure if it's laziness. Judging by the number of sfn and harv errors I've put right, I think it's lack of understanding rather than laziness. I doubt it would enter the mind of the average user to disambiguate dates when adding a list of articles to a "Further reading" section with nothing linking to them. --John B123 (talk) 19:25, 15 June 2020 (UTC)

Break

@Trappist the monk:, can you point to where consensus was reached for this change?

This was the situation before the change. When writing an article, we compile a list of long citations under "Works cited" and use short citations (sfn or harvnb) throughout the text. The short links to the long. If we include a long citation but don't use it, we see a red error message. This is very useful.

What we might do at that point is move the unused reference into Further reading, and remove ref=harv. We do the same when compiling "Selected works" lists of publications. Then if we want to use the same long citation as a source again, we move it back into "Works cited" and add ref=harv; that allows us to restore an sfn citation.

That flexibility has now disappeared, it seems. Either we get ugly error messages all the time in Further reading and Selected works, or we turn them off and don't get the ones we need. SarahSV ^(talk) 03:00, 21 April 2020 (UTC)

You still can all do that, and now you can all do that natively with both CS1 and CS2 templates. If you don't want the warning messages, then don't use scripts that give warning messages. You could also ask for a new script to be developed at WP:SCRIPTREQ which only gives warnings if a citation is in a specific section / omit warnings when in a specific section. Headbomb {t · c · p · b} 03:03, 21 April 2020 (UTC)

As for the discussion, see here and here. Headbomb {t · c · p · b} 03:05, 21 April 2020 (UTC)

Thanks for the links. But we wanted the flexibility. It was a good thing to have templates where ref=harv was the default and others where it wasn't. And we wanted to keep on using the error-message scripts. Someone has made a central change that affects all articles that use templates but without gaining site-wide consensus.

It sometimes feels as though these templates are written for other template writers rather than for article writers. I don't want to be forced to come here to have these discussions, because I have no technical knowledge, so it's hard to take part in them. I don't understand the terms; I'm not sure how to express myself, and there are other things I'd rather be doing. You wouldn't like it if I did something that forced you to spend hours or days discussing the Holocaust. Surely a change that affects all articles using templates should have more consensus. SarahSV ^(talk) 03:16, 21 April 2020 (UTC)

Re: "You could also ask for a new script to be developed at WP:SCRIPTREQ which only gives warnings if a citation is in a specific section". We had one. It relied on the presence of ref=harv if a source wasn't used. How would a new script work? What would it look for? I would like to see this change reverted. It seems to have no benefit and several disadvantages. SarahSV ^(talk) 03:23, 21 April 2020 (UTC)

Ask at WP:SCRIPTREQ and explain which section should have warning and which should not. The errors should always be reported though, given those are actually confirmed problems. Headbomb {t · c · p · b} 03:44, 21 April 2020 (UTC)

What I need is a script that tells me when a long citation isn't used as a source so that I can move it out of "Works cited". We had that, and now it doesn't work because of this change. What is the benefit of the change? SarahSV ^(talk) 04:19, 21 April 2020 (UTC)

What I need is a script that tells me when a long citation isn't used as a source so that I can move it out of "Works cited". Then ask for that at WP:SCRIPTREQ (User:Ucucha/HarvErrors.js still works for that, btw, with warnings outside "Works cited" sections as well). The benefits of this change is that tens of thousands of broken anchors are fixed, and that CS1 and CS2 are now on equal footing and now both work out of the box for {{sfn}}-purposes. Headbomb {t · c · p · b} 04:23, 21 April 2020 (UTC)

These two discussions—here and here—don't amount to consensus. The second is from 2018; the first is more recent but hardly anyone commented. You need strong consensus for a change that affects so many people.

Can you explain about the broken anchors? (Your last point about equal footing means we've lost functionality/flexibility because CS1 and CS2 are now the same; that's not a benefit.) SarahSV ^(talk) 04:53, 21 April 2020 (UTC)

SlimVirgin, can you please link to an article where this change is causing a problem? Thanks. It seems to me that if a source isn't being used in a Works cited section, it will now show a warning (with Ucucha's script in its default configuration) whether that source is using CS1 or CS2 templates, which makes for better consistency for editors.

As for consensus, this page has 365 watchers, and it is the page where changes to CS1 templates are discussed. If you would like to start a new discussion thread about the benefits and drawbacks of potentially reversing this change to CS1 templates, please do so on this page.

Also, all error and warning messages related to this change are hidden from readers and editors unless they have made changes to customize their personal CSS or JS files, so this change affects a tiny fraction of editors, and ways to reduce those effects have been provided. – Jonesey95 (talk) 05:08, 21 April 2020 (UTC)

There is now no way to get rid of these error messages, because removing ref=harv no longer works. The only way not to see them is (a) remove the script, but then we won't have warnings when they are needed in Works cited sections; or (b) rewrite the citations manually for Further reading and Selected works.

Jonesey95, the problem is that it is showing warnings when those templates are used in Further reading or Works cited. So I added a script that Trappist the monk suggested above. But that script has turned off all the warnings, so I don't see the red warnings anymore when long citations are not used as sources. But I want to see those. SarahSV ^(talk) 05:19, 21 April 2020 (UTC)

The script is showing brown warnings for CS1 templates (e.g. cite book) that it has always shown for CS2 templates (e.g. citation). The consensus was that this consistency was an improvement. Another improvement is that the brown warning messages will now show for CS1 templates that are used in Bibliography / Works cited sections but that are not actually cited, and should really be in Further reading sections. If you are willing to link to an article where you see this change as a problem, that would help us understand why the change may be objectionable. (Edited to add: I see that you have included a screen shot from Jedwabne pogrom, where the brown warning messages are helping indicate that two of the full citations are not templated, including one that does not end with a full stop. The warning messages are helping to indicate that CITEVAR improvements could be made.) – Jonesey95 (talk) 05:38, 21 April 2020 (UTC)

"The consensus": can you point to this consensus, please? I think you're mistaken about why the warnings are there; they're there because ref=harv is now the default. SarahSV ^(talk) 05:45, 21 April 2020 (UTC)

It is linked above in response to one of your comments. Do a find on this page for "Yep:". Consistency is specifically mentioned in the first discussion. – Jonesey95 (talk) 06:02, 21 April 2020 (UTC)

Not much of an improvement perhaps, but I have tweaked a copy of HarvErrors.js at User:trappist the monk/HarvErrors.js so that the brown warning messages are not emitted when there are no short-cite templates. In other words, when there are no {{harv}} or {{sfn}} templates, the script is mute so reference sections are not cluttered with the warning messages. I use a different color scheme and my error messages don't shout so the look is a bit different. In User:SlimVirgin/common.js, replace importScript('User:Ucucha/HarvErrors.js'); with importScript('User:Trappist the monk/HarvErrors.js');.

There is now no way to get rid of these error messages, because removing ref=harv no longer works. To quash the script warning messages in a Further reading section (or anywhere else), you can set |ref=none in the cs1|2 template. When set to this value, the cs1|2 templates will not create an anchor ID.

—Trappist the monk (talk) 11:28, 21 April 2020 (UTC)

For the broken anchors, see Help talk:Citation Style 1/Archive 46#broken anchors, I've updated the archive to reflect the old behaviour. For consensus, it's a pretty strong one. This fixes tens of thousands of broken anchors, putting CS1 and CS2 on par with each other, thus allowing for greater flexibility in the choice of CS1 and CS2 styles with an increased user-friendliness. That this reduces the usefulness of an opt-in, optional script used by about 200 active editors is not much of an argument compared to the benefits this brings to everyone else. Headbomb {t · c · p · b} 05:28, 21 April 2020 (UTC)

Headbomb, we've been fine for years without the change. It has broken things. I've posted an image of what I see now in Further reading and Selected works sections. So what I normally do is pull references that I might use in future out of Works cited and into FR, and I remove ref=harv. Then I can easily put them back. What I will have to do now is rewrite them manually so that I (and other editors) don't see these warnings, because removing ref=harv no longer works. SarahSV ^(talk) 05:36, 21 April 2020 (UTC)

The change broke nothing, and while you may have been fine for years without it, articles clearly weren't. And there is a way to get rid of those warnings, don't use User:Ucucha/HarvErrors.js since that's the script that put them there. Again, WP:SCRIPTREQ is there if you want an updated script. Headbomb {t · c · p · b} 05:44, 21 April 2020 (UTC)

I don't see anything broken in the Further reading section at Jedwabne pogrom, except the minor CITEVAR problems that I noted above. You can add |ref=none if you really don't want to see brown warning messages for those citations, but if you later want to move the citations up to Works cited, you'll need to remember to remove it. As it stands, seeing a brown warning for every entry in Further reading is a good way to confirm that those full citations are not being used in the article, as long as they are all formatted consistently.

Also in the Works cited section of Jedwabne pogrom, you might want to disambiguate the three Polonsky & Michlic 2003 citations and their matching short citations, since the short citations don't know which full citation to link to. Same problem with Gross 2003. Let me know if you would like help with any of that. – Jonesey95 (talk) 05:49, 21 April 2020 (UTC)

There two Polonsky & Michlic 2003 in "work cited" that should be trimmed down to one, with the page information moved into the sfn, yes. Also another, which is probably one for a |ref=none since it's not cited directly. The Gross 2003 citation works correctly, there's a collision, but it's not an important one. Headbomb {t · c · p · b} 06:06, 21 April 2020 (UTC)

Re. "You could also ask ... at WP:SCRIPTREQ ..." – anyhow, I posted a place-holder request there (WP:SCRIPTREQ#HarvErrors.js), that is: inviting script-writing co-editors to come here to help determine what is feasible. I was thinking about placing an RfC at WP:VPT to determine whether the "|ref=harv " default is what editors generally want for cs1 cite templates, but didn't do that yet: I am still pondering whether that is a good idea (and if so: how to get it started on the right foot). --Francis Schonken (talk) 07:20, 21 April 2020 (UTC)

Don't invite them here, go there and discuss what you'd want the script to do there. Here is for discussing CS1/CS2 templates, not scripts. Headbomb {t · c · p · b} 10:31, 21 April 2020 (UTC)

"Selected works" section in Coral Lansbury

This is what I see now in "Selected works" and "Further reading" sections that use {{cite book}}, {{cite news}}, etc. Dudley and Gerda, have you been able to find a solution yet? SarahSV ^(talk) 18:28, 21 April 2020 (UTC)

@SlimVirgin: and? You'd see those too if someone used {{citation}} instead. Headbomb {t · c · p · b} 18:36, 21 April 2020 (UTC)

I used to see them if people used {{citation}}, but people usually didn't. If they did, I could switch to {{cite book}} without ref=harv and that fixed it. Now there is no way to fix it. SarahSV ^(talk) 18:46, 21 April 2020 (UTC)

Except switching from CS1 to CS2 based on your preferences is a violation of WP:CITEVAR, and you shouldn't be doing that to begin with. Headbomb {t · c · p · b} 18:56, 21 April 2020 (UTC)

Change can be difficult. In this case, you may find that the new way to eliminate the brown warning messages, adding |ref=none to either {{citation}} or {{cite book}}, is actually easier than converting {{citation}} to {{cite book}}. – Jonesey95 (talk) 18:59, 21 April 2020 (UTC)

I can't tell whether you're being deliberately obtuse. I see these error messages in lots of articles now because more people used {{cite book}} etc than used {{citation}} in FR and Selected works. SarahSV ^(talk) 19:04, 21 April 2020 (UTC)

Nope, deliberately helpful. Things have changed (CS1 templates now work consistently like CS2 templates when it comes to ref=harv, as explained above), and I'm trying to help you adapt to this change. I understand that you are seeing brown warning messages that you did not see before; I have seen them for years when people used {{citation}}, but I have learned to ignore them when they are not indicating an actual error. Adding ref=none to cite book is easy and does not violate WP:CITEVAR, which seems like an improvement to me, although I recognize that it is a different workflow that will take some getting used to. – Jonesey95 (talk) 19:12, 21 April 2020 (UTC)

Sigh. I get that you are frustrated but it is a two-way path. It is equally as exasperating when we offer possible solutions to a problem that are ignored. Read my reply to you. With that tweaked script I do not see any of the warning messages that you are seeing at Coral Lansbury. When I introduce a bogus {{harvnb}} to that article and preview it, I see the appropriate error message and, because the article now has a short cite template (bogus or not), all of the warning messages. Read my reply. Try it. Report back here.

—Trappist the monk (talk) 19:06, 21 April 2020 (UTC)

Not frustrated, Trappist, but upset. I was in the middle of reading for an article I'm writing. Now I've lost my train of thought, and lost time, because of this. These central changes affect a lot of people, and those of us without technical knowledge are lost. As I said before, please imagine that I do something on WP that forces you to discuss the Holocaust in Danish for the next few days, and if you don't, then articles are full of error messages. The only way to get rid of them is for you to discuss the Holocaust in Danish. Imagine that, and you'll understand how I feel. I will now try your script. SarahSV ^(talk) 19:12, 21 April 2020 (UTC)

These changes affect at most ~200 ish active editors who use User:Ucucha/HarvErrors.js, many of which will welcome the added CS1 support. And they will repair tens of thousands of anchors all across Wikipedia for millions of readers. To compare this to the "discussing the Holocausts in Danish" is delusional at best. Headbomb {t · c · p · b} 19:17, 21 April 2020 (UTC)

Trappist the monk, I just tried it. It did the same as the first script you suggested. It removes the brown error messages in FR and Selected works, yes, but it also removes the error messages we want to see, namely when we use a long citation without a corresponding short one. That's one of the points of User:Ucucha/Errors.js. We need the flexibility of being able to add and remove ref=harv. SarahSV ^(talk) 19:22, 21 April 2020 (UTC)

@SlimVirgin: re namely when we use a long citation without a corresponding short one: that is not, and never has been, an error. Most Wikipedia articles are formatted with long citations that do not have short forms referring to them. Perhaps you have something more specific in mind, like "when we add a long citation to a references section that is only supposed to contain things that are referred to by short footnotes". The script you use did catch those, but caught a lot of non-errors besidea. If you want to catch these you will need a smarter script, one that is capable of distinguishing articles that use short citations for all references from articles that use them only for some or for none of the references. —David Eppstein (talk) 19:45, 21 April 2020 (UTC)

SarahSV The degraded functionality of cite book is deliberate and they (whoever they are) are happy with it. They have stopped ref=harv working and made cite book work like citation, in order to make cite book still work even if editors have not formatted it properly. The effect (as I think you are saying) is that you will get an error message if a citation does not link to a source, but not the other way round. Sources which do not link to citations (for example because someone removed citations to an unreliable source but forgot to remove the source from the bibliography) will not give an error message. Apparently the idea is that helping editors who cannot be bothered to format properly is more important than providing full functionality for conscientious editors. Dudley Miles (talk) 19:32, 21 April 2020 (UTC)

Again, |ref=harv works exactly as before, and there is no degradation of anything. It simply has been made default to make templates more user friendly and to repair tens of thousands of broken anchors. However you dealt with warnings in CS2 before still apply here. There is literally no difference in how CS1 and CS2 templates behave. For the millionth time now, if you don't like how User:Ucucha/HarvErrors.js behaves, then make a WP:SCRIPTREQ for a new script. Headbomb {t · c · p · b} 19:34, 21 April 2020 (UTC)

{{Citation}} isn't used as much as FR and Selected works, so I didn't see the brown warnings often. When I did, I sometimes stopped to switch to cite book and that fixed it. But now I see the warnings in lots of articles. Part of what's distressing about this discussion is you're forcing us to explain the same thing over and over and over. We've lost functionality, however you want to describe it. For no good reason and without warning, and without putting a script in place that would solve the problem. I know nothing about coding. But I should imagine that part of being a good coder would be to anticipate problems, find solutions in advance, prepare users for the change, and make the change as easy as possible, with clear explanations and some empathy. None of that happened here. SarahSV ^(talk) 20:18, 21 April 2020 (UTC)

"Part of what's distressing about this discussion is you're forcing us to explain the same thing over and over and over." Well maybe if you listened and actually went to WP:SCRIPTREQ like has been asked of you a million times you wouldn't be so distressed. Or actually followed the hours of advice dispensed. Or answered our questions when we're trying to figure out what it exactly is that you want and consider problematic about the current version except for "it's different, i don't like it". Headbomb {t · c · p · b} 20:27, 21 April 2020 (UTC)

You underestimate how distressing it is to be forced to interact with this level of aggression on the page. I can't walk away from it because then I'm left with the problem. SarahSV ^(talk) 20:46, 21 April 2020 (UTC)

Then perhaps you should re-evaluate how you interact with people then. You're given tons of resources to address your problem, but refuses to make use of any of them. You're pointed to go to WP:SCRIPTREQ, but refuse to do so. No one can help you if you don't let yourself be helped. Headbomb {t · c · p · b} 20:51, 21 April 2020 (UTC)

Are you sure you wrote what it is that you wanted to write? The brown warning messages apply to long citation without a corresponding short one; the error messages apply to a short cite without a corresponding long citation.

I just looked at User:SlimVirgin/common.js and its history.

1. you removed window.checkLinksToCitations = false; – ok
2. you
   1. removed importScript('User:Ucucha/HarvErrors.js'); – ok
   2. added 'User:Trappist the monk/HarvErrors.js' –

The script won't work if it doesn't get imported. Try again? I can't do it for you:

1. remove:

   importScript('User:Ucucha/HarvErrors.js');

2. add:

   importScript('User:Trappist the monk/HarvErrors.js');

—Trappist the monk (talk) 19:42, 21 April 2020 (UTC)

Thank you. I will try again. No, I'm not sure of anything. That's what I'm trying to explain about this being like me forcing you to discuss the Holocaust in Danish. You're speaking a language I don't understand, discussing issues I don't understand, and yet I have to deal with it. It's upsetting. SarahSV ^(talk) 20:22, 21 April 2020 (UTC)

Trappist the monk, it produces error messages where they are wanted (e.g. a long citation without sfn or harvnb in the text), but it also produces them in FR where there is no ref=harv. SarahSV ^(talk) 20:49, 21 April 2020 (UTC)

The addition of ref=harv to cite book was the way linked and unlinked sources were distinguished. Now that ref=harv has been made non-functional, there is no way a script can distinguish between cases where cite book is being used with anchors or intentionally without. Dudley Miles (talk) 20:00, 21 April 2020 (UTC)

Again I ask, how did you deal with this in CS2 before? |ref=harv works exactly as it did before and generates a harvard anchor. Headbomb {t · c · p · b} 20:02, 21 April 2020 (UTC)

Not true. |ref=none inhibits anchor ID creation. This functionality was added several years ago specifically for {{citation}} so that editors didn't have to see the warning messages emitted by HarvErrors.js.

—Trappist the monk (talk) 20:05, 21 April 2020 (UTC)

Bot task related ideas

Combining with Headbomb's suggestion at User talk:Citation bot/Archive_20#remove ref=harv, I see (for the time being) some possible tasks (please add more ideas worth considering if you think of any):

1. remove redundant "|ref=harv" from cite templates:

   Would be a WP:COSMETICBOT task, and would thus require explicit permission for the task before proceeding. Additionally it might make the work harder for those who try to clean up articles by listing footnote-used citations separately from those belonging in "External links"/"Bibliography"/"Works"/"External links" type of sections (since the HarvErrors script seems no longer very useful for such endeavours). --Francis Schonken (talk) 08:34, 19 April 2020 (UTC)

   It would be a cosmetic task indeed (if done alone) with no effect on anything, but it would not make the work any harder for anyone. So a bad task for a bot, if that's the only thing to be done. Headbomb {t · c · p · b} 10:55, 19 April 2020 (UTC)

   Might be worth getting it added to ProveIt's cleanup routines and to WP:GENFIXES though. --AntiCompositeNumber (talk) 02:57, 21 April 2020 (UTC)

   Absolutely yes. Headbomb {t · c · p · b} 10:53, 21 April 2020 (UTC)

   As things currently stand this isn't a WP:COSMETICBOT task - |ref=harv currently throws a maintenance category. Maybe it shouldn't, and maybe a bot removal isn't a good idea, but removing the parameters is not a cosmetic edit according to the definition in the policy. Jo-Jo Eumerus (talk) 19:05, 21 April 2020 (UTC)

   JCW-CleanerBot (talk · contribs) is currently removing |ref=harvs as a stand-alone task (example) – are we OK with that? --Francis Schonken (talk) 18:50, 29 April 2020 (UTC)

   That one is my bad, that should have been made from my account not the bot account. I was testing an AWB pluggin (mentioned at WT:AWB#Running into a limit making a list) to fetch the templates, and forgot to switch back to my account afterwards. The edits themselves aren't problematic though and were all reviewed manually, so no needs to revert. Headbomb {t · c · p · b} 19:00, 29 April 2020 (UTC)

   That template is not suitable for experimentation, so it will be reverted: please take your experiments to pages where you are the main contributor to the content, tx. --Francis Schonken (talk) 21:04, 29 April 2020 (UTC)

   There is no experimentation with the template. There was an experimentation with the AWB / Mediawiki API. Headbomb {t · c · p · b} 21:21, 29 April 2020 (UTC)

2. add "|ref=none" to those templates which currently have no |ref= definition:

   See also above. Would be WP:COSMETICBOT except for those readers who use the HarvErrors script (presumably a minority of readers), thus would need explicit permission too. Apart from that, would probably generate a lot of false positives (e.g., cite templates called by a footnoted reference for which no "ref=harv" was placed in the past). --Francis Schonken (talk) 08:34, 19 April 2020 (UTC)

   This would not have consensus, because the entire point is that having |ref=harv being the default is that thousands of citation notes are fixed, and that it makes the entire citation ecosystem more user-friendly. The solution is to use a script that's up to date. Headbomb {t · c · p · b} 10:55, 19 April 2020 (UTC)

3. define "|ref=none" for all cite templates in "External links"/"Bibliography"/"Works"/"External links" type of sections, unless when the template has a |ref= definition.

   Might generate false positives (e.g. articles erroneously using a "Bibliography" list as list of works used in references). --Francis Schonken (talk) 08:34, 19 April 2020 (UTC)

   See above. Whatever false positives there are far fewer in number than the broken footnotes that it fixes. Headbomb {t · c · p · b} 10:55, 19 April 2020 (UTC)

4. move cite templates which are not called by footnotes to a "Further reading" section if they're not yet in a "External links"/"Bibliography"/"Works"/"External links" type of section.

   Technically more complicated (probably more than Citation bot can do), but might be worth to consider writing such script/code. --Francis Schonken (talk) 08:34, 19 April 2020 (UTC)

   Where to put citations is irrelevant and unrelated to citation anchors being enabled. Headbomb {t · c · p · b} 10:55, 19 April 2020 (UTC)

   The way I understand it, the HarvErrors script was used for script-assisted clean–up of articles. Namely, among other things, differentiate cite template formatted citations used as references from cite template formatted entries not belonging in a "Works cited" type of section. Which is useful clean-up. If the script is no longer useful in assisting such clean-up (which seems a consequence of the "ref=harv" default), and applying repairs and workarounds does not make it useful again for that purpose (or would require more effort than writing some code from scratch), we might consider programming a bot to do such task. Which makes Headbomb's comment above irrelevant to this 4th suggestion. --Francis Schonken (talk) 15:28, 19 April 2020 (UTC)

   The error scripts are still useful for the task described above. If you have five full citation templates in a "Works cited" section, and only four of them are connected to {{sfn}} templates, the unused one will show a brown warning message. See this version of Coniston railway station (England) for an example. Only one full citation in the Sources section shows the brown warning message. In this case, it is because the sfn template that is trying to use it omits the year. – Jonesey95 (talk) 17:33, 19 April 2020 (UTC)

   This one is a horrible idea. For one thing it imposes some bot-enforced ideal of how citations should be formatted, violating WP:CITEVAR. For another it will cause damage to the (probably many) articles that in some way use unlinked references to separately-formatted citations, whether inline or in footnotes, whether formatted as proper parenthetical citations or as prose. —David Eppstein (talk) 05:02, 21 April 2020 (UTC)

Ideas related to getting the HarvErrors script back on track

Two solutions work for me to stop the wrong error message when cite book is used without harv referencing. 1. Adding "window.checkLinksToCitations = false;" to "importScript('User:Ucucha/HarvErrors.js');" 2. Replacing "importScript('User:Ucucha/HarvErrors.js');" with "importScript(David Eppstein/HarvErrors.js);".

However, I have found a new problem. When cite book is used with ref=harv, I am getting an error message "CS1 maint: ref=harv (link)". Based on the comment of Trappist the monk above, this appears to be because CS1 has been changed to make harv referencing the default for cite templates, and give an error message for ref=harv on the ground that it is now redundant. Can I get rid of the wrong "CS1 maint: ref=harv (link)" error message by changing from CS1 to CS2, and if so how do I do that? Dudley Miles (talk) 09:30, 19 April 2020 (UTC)

CS1 maint: ref=harv (link) is not an error message and must not be construed to be one. It is, as it claims to be, a maintenance message. Because |ref=harv is now the default, use of that parameter value no longer has meaning. The module suite adds the category so that editors and their tools can know where these parameters are being used so that they can be removed or modified.

Changing from cs1 to cs2 will not remove the maintenance message if the cs2 template retains |ref=harv. To remove the message, remove the parameter or change its assigned value to something more appropriate.

—Trappist the monk (talk) 11:11, 19 April 2020 (UTC)

When ref=harv was working, if it was in a citation and someone deleted the citation(s) for a book and forgot to delete the book from the bibliography, there would be an error message on the book in the bibliography. If you edited an article which does not use harv, then in order to comply with the article style it was good practice to use cite book without ref=harv, and then you would not get an error message. Now that ref=harv is disabled you presumably either have to get false error messages or not get warnings of errors. So it is now best to turn off error checking entirely because it is a waste of time. Can't we just go back to the system that was working fine? It seems to be Wikipedia's motto now, if something works, break it. Dudley Miles (talk) 11:50, 19 April 2020 (UTC)

ref=harv is still working. And you can still have those warning enabled by using Ucucha as is. What's now happening is that tens of thousands of broken citations now work because 'the system' didn't work as well as it should have. Failing to remove an "unused" citation is not a high-priority issue, because those will still be relevant as general references to articles. And you can always surpess anchors by using |ref=none, just as you could in CS2 before. Now people are free to use either CS1 or CS2, as they should have been a long time ago. Headbomb {t · c · p · b} 12:02, 19 April 2020 (UTC)

Trappist the monk says "Because |ref=harv is now the default, use of that parameter value no longer has meaning." Headbomb says "ref=harv is still working." Which is correct? What are the tens of thousands of broken citations which now work? Who decided that "Failing to remove an "unused" citation is not a high-priority issue"? In some cases they will have been removed because they are unreliable sources. The solution is to stop using citation templates. The latest changes have made them too difficult to use for anyone who is not a technical expert. Dudley Miles (talk) 12:48, 19 April 2020 (UTC)

"ref=harv" generates anchors automatically. This is the default setting now, which means that ref=harv and not putting ref=harv both generate anchors. Nothing broke, just now adding ref=harv is no longer needed because that's done automatically. Stopping to use citation templates would not make any difference here. Anything you wanted to remove, you'd still have to remove manually. You'd lack warnings. Citations would diverge in style and be harder to maintain. Bots couldn't maintain them. The solution here is to update your script, not break thousands of citations. Headbomb {t · c · p · b} 13:00, 19 April 2020 (UTC)

Both are correct. Writing a cs1 template with |ref=harv is the same as writing a cs1 template without |ref=harv. In both cases, the template does the same thing. ref=harv is still working because it is the same as the default. All other values assigned to |ref= are still handled in exactly the same way that they were handled before the module suite update.

—Trappist the monk (talk) 13:36, 19 April 2020 (UTC)

Updating the script would not help with the many thousand articles already written which rely on cite book only generating error messages if ref=harv is used. I would prefer not to have so called "error checking" which gives irritating "maintenance messages" on any article I read which uses ref=harv, and does not give error messages on unused sources. Dudley Miles (talk) 15:04, 19 April 2020 (UTC)

If you do not wish to see the maint category messaging, remove this line from User:Dudley Miles/common.css:

.citation-comment {display: inline !important;} /* show all Citation Style 1 error messages */

Error messages () will still display (except which is currently hidden).

—Trappist the monk (talk) 15:16, 19 April 2020 (UTC)

Thanks. I will try that. Dudley Miles (talk) 15:36, 19 April 2020 (UTC)

But what about all the other people who will see them? This reminds me of the unnecessary "dead-url=yes" change. There are red error messages littering articles all over the place. SarahSV ^(talk) 03:10, 21 April 2020 (UTC)

Again, you are free to not use warning scripts and restrict yourself to error-checking scripts. Headbomb {t · c · p · b} 15:18, 19 April 2020 (UTC)

Whether making articles conform to WP:V is handling of "errors" or handling of "warnings" is an irrelevant distinction. --Francis Schonken (talk) 15:34, 19 April 2020 (UTC)

This not making articles conform to WP:V, this is making articles conform to optional, opt-in, and outdated scripts' bad outputs. Headbomb {t · c · p · b} 11:07, 20 April 2020 (UTC)

IMHO you're losing perspective here. Repairing citations so that a Wikipedia article conforms to WP:V, for which HarvErrors is a designated tool, is far preferable to doing COSMETICBOT edits which, as such, have no relation whatsoever to making articles more or less compliant to core content policies. If the HarvErrors scripts deserve all the epithets you suggest, then go make them better – complaining about them in general is neither here nor there. --Francis Schonken (talk) 11:45, 20 April 2020 (UTC)

Who has proposed to unleash cosmetic bots here? Actual errors are still flagged by all relevant scripts, including User:Ucucha/HarvErrors.js if you like warnings, or User:Svick/HarvErrors.js if you don't. The "fixed" scripts already exist. Headbomb {t · c · p · b} 12:06, 20 April 2020 (UTC)

Re. "Who has proposed ... ?" You proposed a COSMETICBOT task here

Please make up your mind: if HarvErrors is fixed, so that its WP:V-improving assistance is fully operational, it no longer needs the derogatory epithets, does it? --Francis Schonken (talk) 12:22, 20 April 2020 (UTC)

"You proposed a COSMETICBOT task" I did not. I specifically said that removing |ref=harv is not an edit that should be done on its own. Headbomb {t · c · p · b} 12:26, 20 April 2020 (UTC)

... which is still a COSMETICBOT task (an "allowed" COSMETICBOT task does not make a non-COSMETICBOT task), and thus far less substantial than making articles WP:V-compliant, for which a useful script can of course play a significant role. As said, you seem to be losing perspective. I mean, of course, over-all perspective on what best helps the encyclopedia forward. --Francis Schonken (talk) 12:34, 20 April 2020 (UTC)

In WP-linguo, a cosmetic bot is a bot that makes cosmetic edits. This wouldn't be a cosmetic bot task, because the bot would be making substantive changes, and the removal of |ref=harv would be bundled among those. Headbomb {t · c · p · b} 12:42, 20 April 2020 (UTC)

Let's link this to policy: a WP:COSMETICBOT task, i.e. a bot performing cosmetic edits, is usually allowed if the WP:COSMETICBOT task is performed, in the same edit, with a substantial task. That does not make a WP:COSMETICBOT task a substantial task. Aligning citations with core content policies are, in contrast, substantial edits. A script that significantly assists with that is more important than whether or not a substantial edit has an appended cosmetic edit. The loss of perspective seems complete if you think that a cosmetic edit appended to a substantial edit is remotely as substantial as getting a script which has proven its usefulness for articles' WP:V compliance back on track. Trying to make policies say what they don't say is further illustration of that loss of perspective. --Francis Schonken (talk) 13:16, 20 April 2020 (UTC)

This discussion is getting more and more pointless, since you do not seem to understand even the basics of what's at play here. Headbomb {t · c · p · b} 13:20, 20 April 2020 (UTC)

Well, I suppose you needed a reality check one way or another. The basics is WP:V, not bots. --Francis Schonken (talk) 13:23, 20 April 2020 (UTC)

Come back to this when you understand what's going on. Headbomb {t · c · p · b} 13:26, 20 April 2020 (UTC)

My edits above show that I know what's going on, and where not I'm not afraid to ask. Your self-contradicting derogatory language about the HarvErrors script (and its users), your explanations on cosmetic edits (contradicted by policy), etc. were however of little help to better understand what is going on. --Francis Schonken (talk) 13:48, 20 April 2020 (UTC)

Let me educate you then, since you still haven't got a clue about anything going on here. What happened here is that automatically-generated anchors (which were formerly and still are generated by |ref=harv) is now the default behaviour in CS1 templates, just like it was the default behaviour in CS2 templates. This has 2 main effects.

• The first effect is that this fixes tens of thousands of missing anchors, which {{sfn}}-like templates assumed were being emitted, but weren't because someone used a CS1 template instead of CS2 template because they erroneously assumed those templates had the same behaviour. Additionally editors now have a vastly increased flexibility in the choice of CS1/CS2 styles without having to append |ref=harv to CS1 templates (or adding pointless |ref=harv to CS2 templates because this was required in CS1 before). The number of actual errors (namely missing anchors) flagged by users scripts, like User:Ucucha/HarvErrors.js has gone done significantly, leaving only things that now actually require human review, alongside a slight increase in collisions (which are flagged by HarvErrors scripts and other opt-in messages).
• The second effect is that opt-in scripts which "warn" against "missing" {{sfn}}-generated anchors based on CS2-generated anchors, or CS1-generated anchors from |ref=harv, will now warn against "missing" {{sfn}}-generated anchors regardless of which of CS1 or CS2 templates are used. Additionally an opt-in CS1/CS2 maintenance message will display for those that choose to have those enabled. If those warnings bother you, then use User:Svick/HarvErrors.js instead of User:Ucucha/HarvErrors.js, and don't enable opt-in CS1/CS2 maintenance messages.

Finally, the feature request at User talk:Citation bot has nothing to do with WP:V. What it aims to do is to additional cleanup of citation clutter when the bot would make an edit anyway. This is not a cosmetic task, nor would it violate any aspect of WP:COSMETICBOT, because those would be done alongside substantive changes, and would affect absolutely nothing in how the HarvErrors scripts behave, both theoretically and practically. However, it would, over time, remove several of those opt-in CS1/CS2 maintenance messages, as well as reduce edit window citation clutter. Headbomb {t · c · p · b} 14:13, 20 April 2020 (UTC)

I have checked the three solutions suggested and I see that they all work, but with reduced functionality as they do not highlight unused sources. This is an important issue as in some cases citations have been deleted because the sources are not reliable, and in other cases the book should be in further reading to signal to readers that it may contain additional information not covered in the article. I hope this deficiency will be resolved in the future. Dudley Miles (talk) 14:58, 20 April 2020 (UTC)

If you want the warnings, you can still use User:Ucucha/HarvErrors.js as before. Headbomb {t · c · p · b} 15:11, 20 April 2020 (UTC)

That does not help because the rare correct warning will be get lost among the countless false ones which did not give warnings when cite book was used without ref=harv under the previous system. Dudley Miles (talk) 15:50, 20 April 2020 (UTC)

That's really no different than before in an article that used {{citation}} all over without using {{sfn}} and similar templates. Headbomb {t · c · p · b} 16:59, 20 April 2020 (UTC)

Exactly. That is why citation was rarely used in recent edits. When I saw it in an article I would change it to cite book, which was fine until the latest changes degraded its functionality. Dudley Miles (talk) 17:12, 20 April 2020 (UTC)

{{citation}} is used on about 300,000 pages. That's not exactly rare. Headbomb {t · c · p · b} 17:26, 20 April 2020 (UTC)

Also, changing it to cite book etc (on an article that uses citation templates consistently) is a violation of WP:CITEVAR. —David Eppstein (talk) 17:38, 20 April 2020 (UTC)

No, it isn't a violation of CITEVAR. It's a violation if templates are added to an article that doesn't use them, and it's a violation to change the rendered style. SarahSV ^(talk) 19:02, 21 April 2020 (UTC)

Which is exactly what changing CS1 to CS2 does. It changes the rendered style, and thus violates WP:CITEVAR. Headbomb {t · c · p · b} 19:04, 21 April 2020 (UTC)

How many of the 300,000 are with harv referencing? If they are, then there is no problem. They would not have given an error message and I would not have known it is being used. I was referring to the small minority of articles which showed error messages because they use the citation template in the bibliography without harv referencing. In most cases, the template is not used consistently, but as one of several styles. In such cases, I assumed it was OK to change them, but whether I was right or not it will no longer apply. Dudley Miles (talk) 09:22, 21 April 2020 (UTC)

How many? Well, about 142,000 articles use shortened footnotes, so that's a ball park. But many of those are used for CS1 templates. Point is, the vast majority of articles using CS2 templates (at least 285,000 – 142,000 = 143,000) had those warnings on before. Headbomb {t · c · p · b} 10:24, 21 April 2020 (UTC)

@Dudley Miles: there is absolutely no requirement that CS2 be used only in conjunction with harv referencing. CS2 is a valid citation style when used in non-harv footnotes, just like CS1 (and just like CS1 is a valid style even when used with harv referencing). I repeat, if you are changing CS2 to CS1, or vice versa, in consistently-formatted articles, you are in violation of WP:CITEVAR, regardless of whether there is any harv referencing in the article. Don't do it. —David Eppstein (talk) 18:59, 21 April 2020 (UTC)

"Ideas related to getting the HarvErrors script back on track" - how about putting it back how it was before the changes which were plainly made without consensus, and are certainly (judging by all the above) against consensus now? The hundreds of thousands of article-problems created would be seen from any rational standpoint as a powerful argument for reversion. Chiswick Chap (talk) 13:19, 21 April 2020 (UTC)

@Chiswick Chap: The solution is to update your scripts, not re-break citation templates. Headbomb {t · c · p · b} 13:29, 21 April 2020 (UTC)

No, the solution is to abide by consensus. Chiswick Chap (talk) 13:30, 21 April 2020 (UTC)

Consensus is that issues caused by scripts should be resolved by updating scripts. Readers are more important than editors. Headbomb {t · c · p · b} 13:43, 21 April 2020 (UTC)

No, functionality is more important as it benefits editors who can then make correct articles which benefit readers. I'm not claiming importance and nor is anyone else who's objected to these drastic and non-agreed changes here. Chiswick Chap (talk) 15:18, 21 April 2020 (UTC)

Indeed, and the functionality here is to provide anchors for {{sfn}} templates, not provide validation for scripts. Headbomb {t · c · p · b} 15:23, 21 April 2020 (UTC)

How does degrading the functionality of cite book benefit readers? So far as I can see the main effect of the new script is to make it easier for unreliable sources to pass unnoticed in bibliographies. Dudley Miles (talk) 15:40, 21 April 2020 (UTC)

The cite book functionality has not be degraded, it has been upgraded. You can now use harvard/snf templates with it without having to worry about explicitly declaring |ref=harv. See for example Smith (2016).

• Smith, John (2016). Book of Stuff. Oxford University Press.

This fixed tens of thousand of broken footnotes everywhere on Wikipedia. It also makes the conversion of "manual" footnotes like <ref>Smith 2016</ref> very much easier to do without having to worry about finding where the corresponding citation is in the text. Headbomb {t · c · p · b} 15:44, 21 April 2020 (UTC)

It is clearly not true that the cite book functionality has not been degraded. What you have done is to degrade the functionality in order to make the template work even when an editor does not bother to format it correctly. Dudley Miles (talk) 16:19, 21 April 2020 (UTC)

Yes, just like {{citation}} if someone doesn't "bother to format it correctly". That's the point of automatic anchor generation. It's simpler for everyone to use. Headbomb {t · c · p · b} 18:12, 21 April 2020 (UTC)

But that was the disadvantage of {{citation}}. Now you've lumbered the other templates with the same lack of flexibility. Not all template use is for sourcing. SarahSV ^(talk) 20:07, 21 April 2020 (UTC)

If you could live with this "disadvantaged" in CS2, you can live with this "disadvantage" in CS1. That's why those are warnings, and not errors. They flag potential issues, not actual confirmed issues. Headbomb {t · c · p · b} 20:10, 21 April 2020 (UTC)

More importantly, CS2 isn't a special snowflake style. Both CS1 and CS2 are equally valid everywhere on Wikipedia. Headbomb {t · c · p · b} 20:12, 21 April 2020 (UTC)

Example

I never used HarvErrors.js. I do it all manually. Which can be a painstaking endeavour. Maybe if I'd used the script I'd have addressed this train-wreck of a "References" section (qualifying it WP:OVERREF is an understatement) a long time ago. CITEVAR issues regarding the article have been discussed on its talk page, e.g. here.

So, anyone to take up this challenge, that is the challenge of cleaning up "References", "Further reading" etc of that article, with or without script assistance? I would like to get some feedback on how much difference the script (in its original and/or patched versions) makes in tackling such complex cases, and whether the task would have been easier before the "|ref=harv " default for cs1 templates. Tx. --Francis Schonken (talk) 08:15, 21 April 2020 (UTC)

The only thing that needs "cleaning up" as far as footnotes are concerned is that there is there no "Schemelli 1736" reference. All the other anchors work. As for moving things from "References" to another section based on them being directly referenced by {{sfn}}-like footnotes, you can use User:Ucucha/HarvErrors.js if you want. There's no extra difficulty caused by this update. Headbomb {t · c · p · b} 09:56, 21 April 2020 (UTC)

Also technically {{GroveOnline}} is a CS1 template, and not a CS2 template, so there's a slight inconsistency there, which I've fixed. Headbomb {t · c · p · b} 10:38, 21 April 2020 (UTC)

As I wrote above, "CITEVAR issues regarding the article have been discussed on its talk page, e.g. here." The issue discussed there was to return to an earlier references format for the article, with which its first major contributor was familiar: that would be a referencing system using cs1 cite templates (which may not have been clear from the context, but you could have asked, that's what talk pages are for).

Also, you didn't catch a single citation that was not used as reference (and thus doesn't belong in the references system), but was contributing to the bulky "more-than-overref" effect. That's where the HarvErrors script would have come in, I suppose, to get this article in better shape WP:V-wise. --Francis Schonken (talk) 11:50, 21 April 2020 (UTC)

If you have issues with that article, take it to the article's talk page. However, CS2 is the dominant style used there. I don't care which is used, but either convert all to CS1, or convert all to CS2. HarvErrors makes no difference here as far as WP:V is concerned, because this is a style issue, not a WP:V issue. Headbomb {t · c · p · b} 12:50, 21 April 2020 (UTC)

The harverrors script shows a bunch of brown warning messages in the references section of that article. As far as I can tell, those warnings indicate sources that are not being used, so the script is working as a helpful tool to sort those sources into a Further reading section. Since this article used mostly {{citation}}, the brown errors were there before the CS1 update. In a similar article using mostly CS1 templates, before the CS1 module update, only {{citation}} or CS1 templates explicitly using |ref=harv would have shown the brown warnings. Now, the warnings are consistent whether CS1 or CS2 is used. In a situation like this, this change makes the script even more useful than it once was. Unless I am completely missing the point. – Jonesey95 (talk) 15:23, 21 April 2020 (UTC)

Do we have a resolution to this issue yet? If a source is in the reference/bibliography/etc. section, it ought to be cited in the text, and the presence of warnings there that an anchor is not being used is a useful way to highlight the fact that the source either needs to be removed or shifted to a further reading section. Parsecboy (talk) 11:58, 24 April 2020 (UTC)

I think Francis Schonken is still working on An Wasserflüssen Babylon. If you think that some of the full citations in References should be moved to Further reading, you should probably discuss that on the talk page for that article rather than here. – Jonesey95 (talk) 13:33, 24 April 2020 (UTC)

I'm not talking about that article, I'm talking about the problem that the harverrors script no longer works as intended (i.e., if you implement the changes suggested earlier in this wall of text, it only highlights short cites that have no corresponding anchor). It is very useful, especially when overhauling an old article, as it allows one to see when old references have been entirely replaced in the text. I'm not talking about a specific article, this is a general concept. Parsecboy (talk) 10:49, 25 April 2020 (UTC)

Uchacha's and Svick's HarvErrors.js scripts both work exactly the same as before. Nothing has changed there. What happened here is that now both CS1 and CS2 are supported out of the box, rather than just CS2. This has the real-onwiki effect of fixing tens of thousands of broken link for every reader out there. The "downside", if you want to call it such, is that there's a lot more "warnings", but there's no reason for why CS1 templates shouldn't throw those warnings just as much as CS2 templates do.

The above also lead to a newer and better script, User:Trappist the monk/HarvErrors.js which only throws warning on pages that make use of short footnotes, so those that feel the need to suppress warnings no one else but them see, will only see them on pages where those warnings might theoretically be of use. There's also ideas brewing about suppressing warnings in sections where they are not needed. Headbomb {t · c · p · b} 15:23, 25 April 2020 (UTC)

No, citations that are not used as references don't belong in a section that is called "References" – no preliminary discussion is needed about that, it's just something that needs repairing, a WP:SOFIXIT for which one doesn't need to be an expert on the topic. The only thing that is needed is that one cares about WP:V, and subsidiary guidance such as WP:OVERREF. --Francis Schonken (talk) 14:37, 24 April 2020 (UTC)

Except several of these citations may be used as general references. This is where editorial judgment and presentation comes into play. This is not a CS1/2 template issue that needs to be discussed here. Headbomb {t · c · p · b} 14:40, 24 April 2020 (UTC)

The concept of "general references" (whatever you think it may mean) is unknown to the WP:V policy. On the contrary, "... verifiability ... is satisfied by providing an inline citation to a reliable source that directly supports the contribution"; "... The cited source must clearly support the material as presented in the article. Cite the source clearly and precisely (specifying page, section, or such divisions as may be appropriate)." – As I said above, this is a WP:V issue.

There is no problem moving cite-formatted entries which have no apparent use as reference to a "Further reading" section. If that leaves some material unreferenced, the article should be tagged {{refimprove}} or some such, whether or not these works are listed as "References" (etc) or as "Further reading" (etc). --Francis Schonken (talk) 19:47, 24 April 2020 (UTC)

Good thing there is more to Wikipedia than what WP:V thinks exists. See Wikipedia:Citing_sources#General_references. Headbomb {t · c · p · b} 15:41, 25 April 2020 (UTC)

... which (in a quite confusing and self-contradictory way) mainly describes the disadvantages of so-called general references. So, if references which are indicated by footnotes are mingled with others that don't, in the same section, pushing the ones that have no clear connection to any part of the article body, to a "Further reading" section remains an acceptable first step when cleaning up an article. If that leaves some part of the article unreferenced, appropriate tagging, as described in my previous reply, needs to be done anyway, in whatever section the so-called general references are. Of course, if you have access to these documents, better still to sort the sourcing problems of the article, so that no tags need to be placed. But that doesn't diminish the possibility to sort them out to a "Further reading" section as a generally useful step until issues are addressed. --Francis Schonken (talk) 02:30, 26 April 2020 (UTC)

Indeed; it's why we have templates like {{More footnotes needed}} and {{No footnotes}}. The ability to easily locate references that aren't cited is very useful. Parsecboy (talk) 10:52, 25 April 2020 (UTC)

Removal of ref=harv

There was a discussion at User talk:Citation bot/Archive 20#remove ref=harv about removing "ref=harv", and there was no consensus. Despite this, Headbomb continues to do it with AWB, first from templates now from drafts. This makes it a Wikipedia:Fait accompli, and ignores that retaining ref=harv makes it less of a nuisance to add ref=none (assuming this change to the templates sticks).

It concerns me how these discussions have been split across multiple pages. When I tried to open a central discussion at the PUMP, Kees08 closed it down; see the archived discussion. It means no one is keeping track of the objections and Headbomb can continue as if there is consensus. SarahSV ^(talk) 22:33, 29 April 2020 (UTC)

There is consensus, that 3-4 people who use Ucucha's script and prefer a Wikipedia where there are tens of thousand of broken references for millions of readers instead of updating their scripts is not a sensible objection. Headbomb {t · c · p · b} 00:22, 30 April 2020 (UTC)

Almost every time I write about this issue, here and elsewhere (and perhaps every time), you've posted after me, often with gaslighting or a personal attack. It has made me reluctant to post in these spaces. SarahSV ^(talk) 00:45, 30 April 2020 (UTC)

Quoting SlimVirgin: 1. It concerns me how these discussions have been split across multiple pages. 2. Almost every time I write about this issue, here and elsewhere. You are making it difficult to AGF. Despite your forking of the discussions, we have tried to help you everywhere that you have posted and reposted your questions and comments and objections. – Jonesey95 (talk) 05:15, 30 April 2020 (UTC)

Gaslighting and personal attacks? After painstakingly correcting your misconceptions about what citation templates do, what the HarvErrors scripts do, pointing you dozens of times to WP:SCRIPTREQ on your plethora of discussions forks/WP:FORUMSHOPPING, literally working with you to improve scripts, if anyone is gaslighting and making personal attacks on anyone it's you. So if you're concerned about splitting discussions on multiple pages, then stop splitting them. Headbomb {t · c · p · b} 10:56, 30 April 2020 (UTC)

I concur with SV's comment, because you are also attacking an unidentified 2-3 people with an exaggerated mis-statement about their presumed preferences. I really don't understand your claim, especially because the un-updated scripts work just fine with identifying the broken links, but I suspect the cap fits. I never said I prefer "tens of thousands of broken references", not I believe has anyone, and will fix Ucucha errors if they are in my area of interest. What I said about your original solution was that the cure is worse than the disease: the majority of casual readers who saw a red interpolated message that was (a) usually a false positive and (b) only fixable with expert knowledge, as opposed to a bluelink that goes nowhere but I assume is rarely clicked (I have no data to support that). I look forward to the version that's backed by a whitelist, and to a canonical RFC somewhere. David Brooks (talk) 14:33, 30 April 2020 (UTC)

See, this is why this is hard to take seriously. Exactly zero readers saw any red messages whatsoever. The only people who saw any sort of red messages are log-in editors who chose to install a variant of the HarvErrors.js script (Ucucha's, Svick's, Trappist the monk's), and those who have opted in to display error messages generated by Module:Footnotes. Exactly zero red messages from HarvErrors.js are false positives, they are all confirmed errors. Module:Footnotes is experimental and does generate some false positives red error messages, which is why you need to opt-in to see those. The HarvErrors.js scripts, in all variants, are better here, and all work flawlessly.

HarvErrors.js however, now warns about "missing links" to CS1 templates like it did with CS2 templates before, with yellow/brown warnings. New scripts have been updated and several customization options are available, from displaying all yellow warnings, displaying no yellow warnings, to only displaying yellow-warnings on articles that make use of footnotes, and exclude sections where warnings have a low chance of being useful.

However, millions of readers saw tens of thousands of short footnotes with broken links, without any indication that they were broken until they clicked on them. This change remedies that, and also greatly enhances the flexibility of the citation system for editors, and of the HarvErrors.js scripts to find problems on CS1-based articles. And a minor inconvenience to a handful of editors who refuse to update their scripts and or explain what they consider a "good warning" or a "bad warning" does not overcome the impact this has on millions of readers. Headbomb {t · c · p · b} 16:21, 30 April 2020 (UTC)

I am sorry you could not take me seriously. Apparently I was not clear about the type of "red interpolated message"; in particular the first I saw was a false positive, reported to Trappist (albeit with an alarmist heading) on March 27. Back then, almost all citation-class templates with no error in coding or usage (and hence not reported by Ucucha) were complaining with something like sfn error: no target..., and a pointer to a suppression method that usually didn't work. That was the period I was referring to in my concurrence, and pointing that out in no way suggested I was "prefer a Wikipedia where there are tens of thousand of broken references for millions of readers". Perhaps I was being over-sensitive in assuming I was one of the 3-4 people. I think the error messages will reappear, with false positives on a smaller scale, when the feature is turned back on with a whitelist (I admit I have expressed skepticism about the scalability of that approach, while I have run several large survey queries to help Jonesey). "Worse than the disease" because the change did not fix the "link goes nowhere" error, but it did add an alarming but technical and, again, often false, error message, which IMO would have disturbed general users more than the link-to-null behavior by itself. And I think that will remain true. Believe me, I have been following what's going on. I have absolutely no problem with fixing the links, and I'm happy to see efforts to do it more cleanly. David Brooks (talk) 17:52, 30 April 2020 (UTC)

Those errors were generated by Module:Footnotes for a short time on March 27–28, which I reverted on March 28 after less than a day. Those errors and their associated false positives have nothing to do with |ref=harv being default in the CS1/2 update or the HarvErrors.js script. And I also agree with you that those error messages should not be enabled, at least until the false positive situation is resolved. Which it may not end up being. Headbomb {t · c · p · b} 18:23, 30 April 2020 (UTC)

(edit conflict)

Back then, almost all citation-class templates with no error in coding or usage (and hence not reported by Ucucha) were complaining with something like sfn error: no target..., and a pointer to a suppression method that usually didn't work.

That is rather an exaggeration isn't it? almost all? Not true. The error message that you quote was emitted then as it is now by Module:Footnotes. The error messages were, and still are, usually not from false positives though those exist. When I enabled the error messages on 27 March 2020, the category held about 47.7k pages. With whitelisting, with editor fixes, and with the switch to auto-|ref=harv, the category is now at about 34.2k pages. How much reduction of the category size is due each of these three 'reducers' is not known; probably not knowable. The module is used on about 144k pages so clearly not almost all.

—Trappist the monk (talk) 18:31, 30 April 2020 (UTC)

I may have been somewhat imprecise there; otoh you inverted my meaning. Almost all citation-class templates that provide a custom author and/or date and eventually call CS1/2, although correctly coded, emitted a false positive. And I appreciate your suppressing the messages when that was pointed out. But that's the reverse of saying that almost all error messages were emitted by that class of template, which I didn't claim. Look, we're on the same page when it comes to wanting to get the citation mechanism and its links right. I think we're just visualizing different populations of readers/editors and I had misgivings about the direction of this solution which have, I'll admit, have migrated from "serious" to "still nervous". David Brooks (talk) 20:16, 30 April 2020 (UTC)

To be fair to the editors who did this change, having to append |ref=harv whenever I wanted to use sfn templates was also quite irritating. And I am willing to bet that a lot more templates needed that parameter back then than need |ref=none, meaning that the amount of work needed now is less than before the change.

I think we might want to discuss a better way to socialize CS1/2 changes. This isn't the first time where a change caused contention. Perhaps notifying folks through The Signpost - if I understand CS1/2 templates correctly, changes to them are implemented in periodic batches and there is a schedule of sorts of when they change - of major changes to CS1/2 will prevent complaints about breaking changes. Jo-Jo Eumerus (talk) 14:57, 1 May 2020 (UTC)

Given that pretty much no template need |ref=none for articles to be functional to readers, that sort of goes without saying that it's a lot less work. Also the Signpost editors don't like these tutorial articles, so don't bother writing them because you'll be wasting your time, because they'll even try to delete things in your userspace if you try to write something for them that they don't like. Headbomb {t · c · p · b} 15:15, 1 May 2020 (UTC)

Nobody has suggested "tutorial articles"; I was just proposing that changes that can hit many pages be reported there. Jo-Jo Eumerus (talk) 13:00, 2 May 2020 (UTC)

Wouldn't waste my time with that either, but YMMW. A mass message to those that use the old Ucucha's script summarizing {{HarvErrors}} would probably be more impactful, I feel. Headbomb {t · c · p · b} 14:37, 2 May 2020 (UTC)

Sure. My idea was more about preventing similar problems in the future. Jo-Jo Eumerus (talk) 15:29, 2 May 2020 (UTC)

Headbomb continues to remove ref=harv. [1] Pinging RexxS because I don't know what else to do at this point. SarahSV ^(talk) 21:44, 7 May 2020 (UTC)

You can do exactly nothing because this does not affect you or anyone else. If you're curious about why I did those specifically, I took a sampling of Category:CS1 maint: ref=harv (starting at 'Zu') and ran genfixes first, then did the removals on that handful because I wanted to know what was the proportion that would be affected by doing removals on articles where genfixes would be done. The answer is about 15%. Now stop your hounding. Headbomb {t · c · p · b} 21:50, 7 May 2020 (UTC)

Problem

Moved to WP:SCRIPTREQ#Problem. Headbomb {t · c · p · b} 12:54, 4 May 2020 (UTC)

Episode

When, and why, did |episode= in {{Cite episode}} become unsupported? It seems counter-intuitive. -- Michael Bednarek (talk) 02:31, 19 April 2020 (UTC)

This edit.

—Trappist the monk (talk) 02:52, 19 April 2020 (UTC)

I'm sure it was reintroduced later as an alias for |number=. When and why was that removed? Was there any attempt to bot-replace it? Now we see |episode= removed without replacement, a clear disimprovement. -- Michael Bednarek (talk) 02:26, 20 April 2020 (UTC)

If |episode= was reintroduced to {{cite episode}}, you should be able to find its reintroduction and then removal in the template's history somewhere. I was not able to find a reintroduction and removal; I only found the initial removal that I linked above.

When this template migrated to Module:Citation/CS1, I left a note in Module:Citation/CS1/Whitelist (this edit – line 134) wondering if |episode= should be made part of {{cite episode}}. Since then I have done nothing with that parameter in the module suite.

—Trappist the monk (talk) 12:02, 20 April 2020 (UTC)

|episode= was globally supported in that whitelist until 18 April 2020. Reading the module code, it seems the parameter is supported at {{Cite serial}}, so why not at {{Cite episode}}? -- Michael Bednarek (talk) 11:31, 25 April 2020 (UTC)

I migrated {{cite episode}} and {{cite serial}} from their {{citation/core}} implementations to their Module:Citation/CS1 implementations within minutes of each other. Because |episode= was required for {{cite serial}}, that parameter is whitelisted. The migrations were intended to be transparent to readers and editors. The new Module:Citation/CS1 implementation of {{cite episode}} reproduced the {{citation/core}} implementation of {{cite episode}} as it existed on the day of the migration. The old implementation did not support |episode=.

I can't tell you why |episode= was replaced with |title= in 2006; I wasn't here then. The editor who made that change hasn't contributed to en.wiki since April 2019 but you could try asking via email.

—Trappist the monk (talk) 12:11, 25 April 2020 (UTC)

Auto-linking titles with free DOIs

Extended content

Hi all, I would like to propose extending the auto-linking mechanism (providing a default value for |url= when |pmc= is available) to also cover |doi= with |doi-access=free. In short: if |url=, |chapter-url= and |pmc= are not set, but |doi= and |doi-access=free are set, the title of the citation would be linked using the DOI. Example With the following code: {{cite journal | author = Hoffman S.J., Lavis J.N., Bennett S. | year = 2009 | title = The Use of Research Evidence in Two International Organizations' Recommendations about Health Systems | journal = Healthcare Policy | volume = 5 | issue = 1| pages = 66–86 | doi = 10.12927/hcpol.2009.21005 | doi-access = free }} You would get the same result as if |url=https://doi.org/10.12927/hcpol.2009.21005 was set: Hoffman S.J., Lavis J.N., Bennett S. (2009). "The Use of Research Evidence in Two International Organizations' Recommendations about Health Systems". Healthcare Policy. 5 (1): 66–86. doi:10.12927/hcpol.2009.21005.{{cite journal}}: CS1 maint: multiple names: authors list (link) Motivation As a reader, it is natural to click on the title of a citation to access it. Clicking on identifiers is less intuitive, even when they are marked as free with . In fact, editors often fill the |url= field with the URL the DOI resolves to, to obtain this linked title (see statistics below). When an article is free to read from the publisher, it is generally preferable to point readers to the publishers' version. Not only is it the version of record, but this is also the place where any errata or retraction will be published, and the DOI link is less likely to rot. If for some reason another version is preferred (because it is the one the editor read when citing the work, or because it is more directly accessible than via the publishers' website), it would still be possible to override the link with |url=, just like for |pmc=. Since |pmc= is already used for auto-linking, I propose that |pmc= has priority over |doi=+|doi-access=free to generate the title link. This will ensure that all titles currently linked will not be changed by this move. PubMedCentral also stores versions of record, in a clean and readable way, so I do not think there is a need to override them. Similar ideas have been suggested before, for instance here by User:Headbomb. Statistics Here are some figures extracted from the enwiki dump of 2020-04-20. At this point, 189,097 citations had a |doi= and |doi-access=free set. 1,773 (1%) of these also had a |pmc= 28,012 (15%) did not have a |pmc= but had a |url= or |chapter-url= 159,312 (84%) had none of |pmc=, |url= or |chapter-url=, so their title was not linked, but would be linked using the DOI with this proposal. Among the templates with both a free-to-read DOI and a manually-specified URL, 66% of these are equivalent to the DOI link (they eventually redirect to the same website) or the URL points to the publishers' website but no longer works (404 error). This figure was obtained by randomly sampling 100 pairs of DOI/URL from the dataset and comparing the links manually. This shows that editors are keen to link the title of their citations. This encourages link rot since they rarely use |url=https://doi.org/... but rather the URL the DOI redirects to. Discussion Do we need to have an RFC about this? I am available to change the Lua code in the sandbox to implement the move if there is consensus for it. − Pintoch (talk) 12:55, 23 April 2020 (UTC) Previous relevant RFCs include: Wikipedia:Village pump (proposals)/Archive 135#Access Locks: Citation Template Behaviour RFC (Aspect B4) Help talk:Citation Style 1/Archive 59#RfC on linking title to PMC I don't think there is consensus for this change. Kanguole 13:14, 23 April 2020 (UTC) Support There was an RFC about this a while back, and there was fairly widespread support for the idea (see B4 above, nearly every comment was in support, I have no idea how the closer got no consensus from this, and most the opposes didn't were self-contradictory or based on minor technical details that can be easily solved). Auto-linking would be absolutely great. The only things that really needs to be decided is two things What's the default hierarchy (e.g. PMC > free DOI > free Bibcode > etc...). This should only apply to garanteed version of records, and not include preprints/general repositories like arXiv and CiteSeerX How do you overrule the default hierarchy (e.g. |auto-url=bibcode, |auto-url=citeseerx, etc...) Headbomb {t · c · p · b} 13:15, 23 April 2020 (UTC) Also I wouldn't say clicking on identifiers is any less intuitive. However auto-linking the title is certainly more accessible. Headbomb {t · c · p · b} 13:18, 23 April 2020 (UTC) My intention was to baby-step this by only adding DOIs first, as I think this is more likely to pass as a RFC (because the DOI normally points to the version of record) and does not require introducing heavy machinery to configure the auto-linking precedence. But I am not opposed with auto-linking other identifiers if there is broad support for it. The risk is repeating the RFC linked above (B4), and proposing an overly complicated system which puts off editors. − Pintoch (talk) 13:26, 23 April 2020 (UTC) I'm entirely fine by starting with DOIs autolinking (with something like PMC > DOI, with |auto-url=doi over-riding the default) and then phasing in the others afterwards. Headbomb {t · c · p · b} 13:39, 23 April 2020 (UTC) There was substantial support in that RFC, but there was also substantial opposition, which would be why that aspect was closed as "no consensus". Kanguole 13:47, 23 April 2020 (UTC) Support the proposal and thank you for the statistics. I think this proposal is consistent with consensus on how to work towards the overarching goal of Wikipedia:Verifiability. There was strong support for adding the PMC parameter and use it for links, I think it was a success. Given recent discussions and requests, I believe it would also appropriate to "linkify" the title when |hdl-access=free is set. For context, this includes many Hathi Trust URLs, like [2] which goes to [3]. These could be treated like PMC (used before the DOI, but after PMC), so as to preserve the links which no longer appear after a few thousand Handle System URLs were converted into |hdl= only. Nemo 13:24, 23 April 2020 (UTC) The hdl discussion which you linked to is as yet unresolved, so brought it back from the archive. If hdl is included, as you propose, that would be a firm oppose from me, as long as no clarity is given where we stand with that discussion. About the doi, I'm mildly positive. The hdl's access status is often difficult to define: it may be "free" when one visits the website from one country, and "non-free" if accessing the website from another (for that reason, I suppose, WorldCat's indications of whether a book is or isn't freely accessible at Hathitrust are notoriously unreliable) – dois have this problem far less in my experience: mostly either it's free for all, or subscription/registration for all. Anyway, bots should stop removing urls if there is a doi (there too, I don't know whether Citation bot or any other bot is still doing that – here's a beauty of one bot inserting doubled links in the same type of citation where another bot was removing somewhat similar redundancies). So there too, although I'm mildly positive for the doi's, the idea would be a firm "no" as long as it can not be ascertained that presence or absence of urls and dois is *as decided by human editors*, not what one bot after another makes of it. --Francis Schonken (talk) 14:10, 23 April 2020 (UTC) You seem to be very confused about just about everything here. There is no "warring bots". Citation bot cleans up a DOI url to a |doi= parameter as it should (much like it cleans up handle links to |hdl=). What GreenCBot does there is adds a free archived version of the PDF. There was an issue with GreenC bot, which has subsequently been fixed. This has nothing to do with Citation bot, or auto-linking free DOIs. Headbomb {t · c · p · b} 14:25, 23 April 2020 (UTC) Oppose as in the above RFC: Links for identifiers are generated anyway, but duplicating them on the title adds nothing but a bit more complexity and blue text. If a reader sees a linked title, is it an explicit |url= or a copy from one of the identifiers? If there is more than one identifier, which URL was copied? Let's avoid all that, by not copying. (I know we already copy links from |pmc=, and I'd like to remove that, but concede that there won't be consensus for that.) Kanguole 13:47, 23 April 2020 (UTC) "If a reader sees a linked title, is it an explicit |url= or a copy from one of the identifiers?" Why does that even matter? The only thing that should matter here is getting the reader to the free version of record. We should be consistent and do it for DOIs as well as PMCs Headbomb {t · c · p · b} 13:51, 23 April 2020 (UTC) I concur - I think most readers do not care about identifers. They should not need to know about the difference between a DOI and a HDL to read a cited article. − Pintoch (talk) 14:00, 23 April 2020 (UTC) It matters because simplicity of interface is important. I am opposed to having this wedge pushed any further in. Kanguole 14:04, 23 April 2020 (UTC) "Click on the title" is about as simple as it gets. Headbomb {t · c · p · b} 14:08, 23 April 2020 (UTC) If we duplicate free DOI URLs as well as PMC ones, the next step will be to duplicate the other free URLs, and then we will need a priority ordering as suggested above, and it will not be a simple matter for a reader to work out what they are going to get if they click on the title. As you say above, clicking on identifiers is simple (particularly when they are marked as free). Kanguole 14:21, 23 April 2020 (UTC) will not be a simple matter for a reader to work out what they are going to get if they click on the title They are going to get a free version of the article. Headbomb {t · c · p · b} 14:28, 23 April 2020 (UTC) Especially if you're on a desktop device and can hover over it, to see what link is actually there. (Don't tell me if you click on links without checking to see whether the alleged link is the real one, okay? That way lies hacked accounts.) I think that the assumption that readers know what those weird numbers at the end of the citation mean is extremely doubtful, especially for the vast majority of readers without graduate degrees. I think it would be ideal to have a link on the title whenever possible, even when that link is a duplicate. WhatamIdoing (talk) 15:17, 23 April 2020 (UTC) My position is the same as in the previous discussion. Yes, you'll need an RFC. No, I don't think here is where you should have it given the even split down the middle last time. WP:MOS or WP:VPPRO or WP:CITE. --Izno (talk) 16:45, 23 April 2020 (UTC) Thanks, I'll start an RFC at WP:VPPRO. − Pintoch (talk) 18:52, 23 April 2020 (UTC)

RFC

@Kanguole, Headbomb, Nemo bis, WhatamIdoing, and Francis Schonken: Here is an RFC: WP:VPPRO#Auto-linking_titles_in_citations_of_works_with_free-to-read_DOIs. − Pintoch (talk) 19:09, 23 April 2020 (UTC)

And I forgot to ping Izno, sorry. − Pintoch (talk) 20:52, 23 April 2020 (UTC)

Is it time to ask someone to close the RfC? Nemo 06:10, 2 May 2020 (UTC)

Early closes of RfCs can be proposed at WP:ANRFC (without much guarantee such early close proposal would be enacted upon). --Francis Schonken (talk) 06:15, 2 May 2020 (UTC)

I have done so. I agree it is early in comparison to the rest of the backlog but the participation has been good already and the consensus seems pretty clear. − Pintoch (talk) 17:45, 2 May 2020 (UTC)

Why? Is this a race? Will the sky fall if the rfc isn't decided now? I think that you should withdraw the early-closure request unless you can show that a normal rfc duration is somehow detrimental to something (what that might be, I don't know).

—Trappist the monk (talk) 17:55, 2 May 2020 (UTC)

There's no such thing as "early closure" of this RfC, which didn't have an advertised end date. "An RfC should last until enough comment has been received that consensus is reached, or until it is apparent it won't be. There is no required minimum or maximum duration" (Wikipedia:Requests_for_comment#Duration). One week is a rather common timescale for a number of decision-making processes on the English Wikipedia, we're not in WP:SNOW territory. Sure, the discussion can continue a few weeks if there's a point to it, but none has been offered so far. Nemo 18:12, 2 May 2020 (UTC)

Legobot appears and adds an {{rfc}} template. Thirty days later, Legobot returns and removes the {{rfc}} template. rfcs held here on this page over the last little while have all ended after Legobot removed the {{rfc}} template. Is there a reason why it is necessary to close the rfc before the thirty days?

—Trappist the monk (talk) 18:49, 2 May 2020 (UTC)

No opinion on whether or not this should be closed, but flipping the question around, is there a point to keeping the RFC open when the outcome is obvious? Headbomb {t · c · p · b} 18:51, 2 May 2020 (UTC)

I'm indifferent. When the norm around here has been to wait for Legobot to remove the {{rfc}} template, it just seems odd to me that there is this push to closure. I'm curious to know why there is such a rush.

—Trappist the monk (talk) 19:13, 2 May 2020 (UTC)

There is no particular rush, just excitement about seeing the citation templates get better! Is it not thrilling when we have an opportunity to deliver a simple change that will make editors and readers' lives easier? Seeing the surge of enthusiasm for the move is definitely motivating me to implement it swiftly. But I will not make the change in the sandbox until the RFC is closed. A WP:SNOW closure is proposed on the RFC, make sure you make yourself known if you oppose this closure (and more generally, your participation in the RFC would be welcome, of course, but I also understand if you prefer to remain neutral). − Pintoch (talk) 16:37, 4 May 2020 (UTC)

Implementation

The RFC has been closed, so I have implemented the change in the sandbox:

Cite journal comparison

Wikitext	{{cite journal|author1=Hoffman S.J.|author2=Lavis J.N.|author3=Bennett S.|doi-access=free|doi=10.12927/hcpol.2009.21005|issue=1|journal=Healthcare Policy|pages=66–86|title=The Use of Research Evidence in Two International Organizations' Recommendations about Health Systems|volume=5|year=2009}}
Live	Hoffman S.J.; Lavis J.N.; Bennett S. (2009). "The Use of Research Evidence in Two International Organizations' Recommendations about Health Systems". Healthcare Policy. 5 (1): 66–86. doi:10.12927/hcpol.2009.21005.
Sandbox	Hoffman S.J.; Lavis J.N.; Bennett S. (2009). "The Use of Research Evidence in Two International Organizations' Recommendations about Health Systems". Healthcare Policy. 5 (1): 66–86. doi:10.12927/hcpol.2009.21005.

Let me know if you spot any issue with the implementation. Thank you to everyone who participated! − Pintoch (talk) 18:32, 22 May 2020 (UTC)

On title-link and other things

It already starts to get nasty...

• The |title-link= parameter is not supported properly. If specified it must override any other settings. Right now (also in the current version), it throws the non-sensical error message: "|pmc= missing title" (or, in the sandbox version, also "|doi= missing title" if |pmc= is not given).
• For cite templates supporting chapters (like with |mode=book), there are a number of corner-cases if |chapter-url= is given instead or in addition to |url=, and with or without |chapter= given in addition to |title=. By default, |title-link= and |url= are for |title=, and |chapter-url= is for |chapter=. However, if |chapter=, |url= and |title-link= are not given, |chapter-url= should be used for |title= as well.
• There is no parameter to define which identifier should be used if multiple are available, or to force the auto-linking feature to be disabled. There are cases where linking to the doi or pmc is undesirable. So, we need something link |auto-link=none/pmc/doi - alternatively, the |title-link= parameter could be used for this as well, like |title-link=none/pmc/doi/... (where ... could be any Wikipedia article name (except for those few tokens used to specify an identifier) and IMO the default should be "none").

--Matthiaspaul (talk) 19:45, 24 May 2020 (UTC)

Just as a clarification to your "non-sensical error message": The title-link parameter is for internal wikilinks on titles, most often used for books for which we have articles. It is an error to try to place both an internal wikilink and an external url link on the same title. The unclear error message that you are seeing is from that error. So I agree with your main point here: autolinking must detect the situation that there is an internal wikilink on a title, just like it detects the situation that there is an explicit url= link, and not try to autolink in that case. It would also be helpful to have a way to disable autolinking more generally, but I think using a link=yes parameter to specify link=no is not exactly an intuitive design. —David Eppstein (talk) 20:00, 24 May 2020 (UTC)

@Matthiaspaul: good point about the issue with |title-link= - that was already an issue with |pmc= actually. I have disabled auto-linking when a |title-link= is provided - I think that is quite consensual? For your second point, could you give some examples of the issues you have noticed using {{cite compare}}? About adding a parameter to disable auto-linking in general, I think the best way forward would be that the RFC participants play by the rules and accept the consensus instead of trying to introduce backdoors in its implementation. You are of course welcome to start a new RFC once this change has been deployed, or challenge the closure of this RFC if you think it was conducted inadequately. Thank you! − Pintoch (talk) 21:44, 24 May 2020 (UTC)

I disagree. Use of |title-link= must not usurp a link to the source.

Auto-linking |title= with |pmc= / |doi= is only supported for {{cite journal}}. {{cite book}} has nothing at all to do with title auto-linking from identifiers; it does not happen.

I do agree that the error message is less than helpful; it is there to catch the legitimate case when |url= is set but |title= is not set. I have sandboxed a correction that will emit the message when |url= and |title-link= are both set. I did not save that because I think that the edit granting |title-link= priority over |url= should be reverted.

—Trappist the monk (talk) 22:12, 24 May 2020 (UTC)

@Trappist the monk: fair enough, I am happy with your solution too! Reverting my change now. − Pintoch (talk) 22:14, 24 May 2020 (UTC)

Here is the fix:

Cite journal comparison

Wikitext	{{cite journal|journal=Journal|pmc=12345|title-link=Title|title=Title}}
Live	"Title". Journal. PMC 12345.
Sandbox	"Title". Journal. PMC 12345.

Cite book comparison

Wikitext	{{cite book|title-link=Title|title=Title|url=//example.com}}
Live	Title. {{cite book}}: URL–wikilink conflict (help)
Sandbox	Title. {{cite book}}: URL–wikilink conflict (help)

—Trappist the monk (talk) 22:48, 24 May 2020 (UTC)

There are notable journal papers for which title-link would be appropriate and for which autolinking should not happen when title-link is set. Grothendieck's Tôhoku paper, as an example (of a notable journal paper, not of one with an open doi). As for autolinking not being available for {{cite book}}: whyever not? Books can have dois and I don't see why those dois might not be open access in some cases. —David Eppstein (talk) 04:01, 25 May 2020 (UTC)

No doubt. Maybe that is why {{citation}} doesn't have auto-linking capability:

{{citation/new|first=A.|last=Grothendieck|authorlink=Alexander Grothendieck|title=Sur quelques points d'algèbre homologique |title-link=Grothendieck's Tôhoku paper |journal=[[Tôhoku Mathematical Journal]]|volume=9|issue=2|series=(2)|pages=119–221|year=1957|mr=0102537|doi=10.2748/tmj/1178244839|doi-access=free}}

Grothendieck, A. (1957), "Sur quelques points d'algèbre homologique", Tôhoku Mathematical Journal, (2), 9 (2): 119–221, doi:10.2748/tmj/1178244839, MR 0102537

The same can be applied to books.

—Trappist the monk (talk) 13:55, 25 May 2020 (UTC)

@Trappist the monk: Just to be sure I understand: how are users supposed to solve the error that arises when both |pmc= and |title-link= are set? Remove one of them? Is that really satisfactory? I do not really see why we would forbid people from wikilinking the title if the PMC link is available as an identifier. Just like |url= can be used to override the link generated by auto-linking, I would expect the same of |link-title=. Of course we are talking about very rare corner cases, but it is worth getting them right. I am also not sure why {{cite book}} would behave differently: this has been the case so far because |pmc= is probably never applicable to books, but as David points out there are plenty of books with DOIs nowadays. − Pintoch (talk) 05:22, 25 May 2020 (UTC)

While I do not support the auto-linking feature myself (because it violates fundamental rules of good user interface design), if it gets implemented, it should be implemented as consistent as possible across all cite templates. Nobody can expect users to understand that it works for {{cite journal}}, but not for {{cite book}} etc. Implementing this differently and addressing the differences in the documentation would add unnecessary complexity without offering any benefit in return.

Regarding priorities, I think it is paramount that |title-link= cannot be overridden by other possible link targets, because internal links have priority over external links. Since there is no way to support a wikilink and an URL link in the same title and we should not silently suppress information given in a cite template, I think it is desirable to display an error message if both are available at the same time. The normal solution would be to remove |title-link= rather than |url= (or, where applicable, |chapter-url=), but it is also possible the other way around but much less likely to happen.

All the other external links from identifiers which could be used for auto-linking should only be used for the title link if |url= (or, where applicable, |chapter-url=) is not given. However, as this is an optional feature, the presence of external links from identifiers should also never override |title-link= and should never generate an error message, if both are present at the same time (after all, external links from identifiers are still present through their identifiers, so there is no display conflict).

Again, there are cases were auto-linking is inappropriate and there must be some way to override it. Otherwise, people would solve the problem by not providing identifier information in the first place, or not using citation templates at all - hardly a good solution. Also, in the discussion and the RFC, people have already stated that PMC and DOI will just be a start and that they plan to add support for auto-linking further identifiers in the future. As multiple identifiers can be given at the same time, we must define auto-linking priorities to them, and since whatever scheme we will come up with will not work for some cases (PMCs are not always "better" than DOIs, etc.), there must be some way to override this automatic selection and deliberately specify a particular identifier to be used for auto-linking. For this, we could introduce a new parameter like |auto-link=, but I suggest to overload the normal functionality of the |title-link= parameter for this by letting the parameter accept a number of special tokens (that is, the keyword "none" and the parameter names of all supported identifiers). If none of these special symbols like "pmc" or "doi" is given, |title-link= is treated like before - defining the internal link. I think, the corner-case that we'd want to link to an article named after an identifier is very rare (and we could still go through a redirect in this case), so no actual conflict would arise out of this double-use of the parameter. If the selected identifier is not provided, this could either be silently ignored or a warning be given.

One more thought: As already said, in contrast to the outcome of the RFC my personal preference for the auto-selection default would be "none", so that it would have to be deliberately enabled where actually needed (and possibly useful). If it would be implemented this way, we would not even need a "none" keyword and also would not have to set up priorities for the auto-selection; |title-link=doi would enable auto-linking for the |doi=, |title-link=pmc for the |pmc=, ..., |title-link=Title would link to Title etc. This would reduce unnecessary complexity, be more predictable (no surprises because the title auto-linking feature is only used when deliberately enabled) and much easier to explain in the documentation:

"If the title link should duplicate one of the links given as identifiers, you can select the desired identifier through |title-link=<identifier-parameter> without having to duplicate the identifier link as |url=".

Clean and easy.

--Matthiaspaul (talk) 12:42, 25 May 2020 (UTC)

because internal links have priority over external links I think that a citation is needed for that. In cs1|2, external links always have priority over internal links. We often see stuff like:

{{cite book |title=[[Abraham Lincoln]] |url=//example.com}}

which, because a single link cannot target multiple locations, produces an error message:

[[Abraham Lincoln]]. {{cite book}}: URL–wikilink conflict (help)

|title-link= is handled in the same way; |url= links the title and the citation emits an error message:

{{cite book/new |title=Abraham Lincoln |title-link=Abraham Lincoln |url=//example.com}}

Abraham Lincoln. {{cite book}}: URL–wikilink conflict (help)

In these cases we do not silently suppress information given in a cite template.

this is an optional feature Ha! For |pmc=, not so (alas). When I disabled the oddity that is {{cite journal}} with |pmc= set and |url= not set (a comment that used to be in the code) I died on that hill in a battle with WP:MED. While I have some sympathy for your |title-link= solution (I would have chosen to add keywords to the various url-holding parameters instead) you too, will die on the hill if you attempt to switch WP:MED from the fully automatic |pmc= to some sort of semi-automatic mechanism. Were we to implement a semi-auto-link mechanism, we will still have the abomination of special-case code because |pmc= will be fully automatic.

—Trappist the monk (talk) 15:07, 25 May 2020 (UTC)

Thanks for correcting me regarding that |url= takes priority over |title-link=. However, the basic argument remains valid that both cannot exist at the same time and the user has to remove one of them to get rid of the error message - which one should be a deliberate decision of the editor.

Regarding auto-selection or not, regardless if auto-selection for auto-linking will be enabled by default (by absense of a |title-link= parameter) or only on demand (through |title-link=), we need a parameter to either override or control the behaviour. If we use something like |title-link=doi (my suggestion) or |url=doi (your suggestion) for this does not matter much, however, I find |title-link= more intuitive to remember for this (and it might cause less confusion for bots trying to make sense of invalid URLs like "doi".).

Regarding the MED folks, what is so difficult for them to add something like |title-link=pmc to a citation, if it makes the auto-linking system as a whole more consistent and easier to understand for the majority of people outside of MED topics?

--Matthiaspaul (talk) 18:04, 30 May 2020 (UTC)

Followup regarding I think, the corner-case that we'd want to link to an article named after an identifier is very rare (and we could still go through a redirect in this case), so no actual conflict would arise out of this double-use of the parameter.

Instead of going through redirects to work around a conflict of a desired target article name with a parameter name for an identifier, we could use the "((just do it))" syntax to enforce the interpretation as article name rather than parameter name. So |title-link=((pmc)) would link the title to pmc instead of retrieving the link target from the |pmc= parameter, same for the other supported identifiers.

--Matthiaspaul (talk) 18:04, 30 May 2020 (UTC)

In the first sentence of the doi auto-link RfC you mention CS1/2 and in the second sentence you mention |chapter-url=. I guess that when you did your research before proposing the doi auto-link RfC that you did not notice that the pmc auto-link applied only to {{cite journal}}, has always applied only to {{cite journal}}. Go look at the pre-lua versions of the templates to see that.

The Grothendieck citation that I mentioned above is a solution for both journal and book cites with |pmc= or free-to-read |doi=. The help text could use a bit of a tweak to suggest this solution (among other things that need fixing there). And, because it is the first error message mentioned, could also use a bit of a tweak.

—Trappist the monk (talk) 13:55, 25 May 2020 (UTC)

@Trappist the monk: Yes indeed, the RFC text implies that the auto-linking would be deployed not just to cite journal, but to other CS1/2 templates as well, otherwise I would not have mentioned |chapter-url=. Yes, so far it applied only to cite journal, but the point of this RFC is to introduce a change in the behaviour of the CS1/2. So I do not see why we should preserve this restriction: even editors who oppose auto-linking agree that it would not make sense to keep it. So: we should enable auto-linking for all CS1/2 which accept doi or pmc, and disable auto-linking when a link-title is supplied. We cannot expect editors to switch to {{citation}} to disable auto-linking - even if we explicitly pointed to this "solution" in the error message: it just does not make sense. I understand you have a great knowledge of the internals and history of these templates and you probably find most of this very natural, but we cannot require editors to study the History of the English Wikipedia Citation Templates in five volumes, so that they know about the migration to Lua, the removal of pmc auto-linking, the WP:MED rebellion, and so on, to make sense of the historical oddities that we preserve for the enjoyment of future generations. Let's just make this usable. Please revert your own change to add the error message, I will then restore my version and generalize the auto-linking beyond {{cite journal}}. − Pintoch (talk) 15:24, 25 May 2020 (UTC)

Can you show how the error message fix that I made is somehow wrong and, because it is wrong, needs to be reverted?

Here are sixteen {{cite journal}} templates that cover the sixteen possible combinations of |pmc=, |title=, |url=, and |title-link=. Where there is an error message, the template is being asked to do something that it cannot do because of insufficient parameters or too many parameters vying for the use of a single resource:

• 0000: Journal. {{cite journal}}: Missing or empty |title= (help) – no pmc, no title, no url, no title-link
• 0001: Journal. {{cite journal}}: Missing or empty |title= (help) – title-link
• 0010: Journal //example.com. {{cite journal}}: Missing or empty |title= (help) – url
• 0011: Journal //example.com. {{cite journal}}: Missing or empty |title= (help) – url, title-link
• 0100: "Title". Journal. – title
• 0101: "Title". Journal. – title, title-link
• 0110: "Title". Journal. – title, url
• 0111: "Title". Journal. {{cite journal}}: URL–wikilink conflict (help) – title, url, title-link
• 1000: Journal. PMC 12345 https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12345. {{cite journal}}: Missing or empty |title= (help) – pmc
• 1001: Journal. PMC 12345. {{cite journal}}: Missing or empty |title= (help) – pmc, title-link
• 1010: Journal. PMC 12345 //example.com. {{cite journal}}: Missing or empty |title= (help) – pmc, url
• 1011: Journal. PMC 12345 //example.com. {{cite journal}}: Missing or empty |title= (help) – pmc, url, title-link
• 1100: "Title". Journal. PMC 12345. – pmc, title,
• 1101: "Title". Journal. PMC 12345. – pmc, title, title-link
• 1110: "Title". Journal. PMC 12345. – pmc, title, url
• 1111: "Title". Journal. PMC 12345. {{cite journal}}: URL–wikilink conflict (help) – pmc, title, url, title-link

Because cs1|2 prioritizes external links over internal links, these templates render the title linked with the value assigned to |url= (1st) or |pmc= (2nd) when in conflict with |title-link= (or with a wikilink embedded in the |title= value). This is consistent for all cs1|2 templates and should not change. Making it so that auto-links from |pmc= and |doi= yield to |title-link= is inconsistent with the current handling of URL–Wikilink conflicts. To make cs1|2 consistent in a way that makes external links yield to internal links would mean that when editors write stuff like:

{{cite book |title=Title with partially wikilinked [[title]] |url=//example.com}}

would render as:

Title with partially wikilinked title.

And that is nonsense.

If this mechanism is to be made available to all cs1|2 templates and if it can be shown that it is absolutely required that auto-links yield to |title-link= then, no doubt, some sort of override mechanism can be concocted. Yes there are en.wiki articles about books and journal articles. I think that it misleads the reader who, seeing the blue-linked title, clicks it expecting to go to the actual book or journal article, and ends up at an en.wiki article about that book or journal article. It has happened to me. |title=, when linked, is an advertisement to see the actual source that is advertised. That was the whole purpose of this RfC was it not? Click this title to read the source.

—Trappist the monk (talk) 18:08, 25 May 2020 (UTC)

@Trappist the monk: Can you show how the error message fix that I made is somehow wrong and, because it is wrong, needs to be reverted yes, I have done so repeatedly, and so have David Eppstein and Matthiaspaul. You understand the problem very well, you are just trying to instrument this corner case to force the introduction of a parameter to disable auto-linking, because the outcome of the RFC is not to your taste. Statements such as "|title=, when linked, is an advertisement to see the actual source that is advertised" are obviously wrong: in that case, using |title-link= to point to Wikipedia articles should be forbidden, as they violate this rule. You are well aware of the fact that internal and external links are not rendered the same way, see the difference between your cases 0101 and 0110. I will explain one last time what the issue is: raising an error when pmc, title and title-link are present (your 1101 case) is wrong: in this case, the |title-link= should be used to link the title, without raising an error. If an editor uses |title-link= in that situation, they want to use this link on the title, and we need to respect this choice. The auto-linked URL is only a default value, that can be overridden by editors with any external or internal link. I have reverted my own proposal to let you try yours, which does not suit anybody except you: now is the time to accept that. Thank you in advance for your cooperation and good faith. − Pintoch (talk) 18:36, 25 May 2020 (UTC)

I'm pretty sure that you do not understand what it is that my fixed. So, direct comparisons: old v. new from the examples above. Remember what I said: this is an ; nothing more:

• 0111 (old): "Title". Journal. {{cite journal}}: URL–wikilink conflict (help) – title, url, title-link
• 0111 (new): "Title". Journal. {{cite journal}}: URL–wikilink conflict (help) – title, url, title-link
• 1101 (old): "Title". Journal. PMC 12345. – pmc, title, title-link
• 1101 (new): "Title". Journal. PMC 12345. – pmc, title, title-link
• 1111 (old): "Title". Journal. PMC 12345. {{cite journal}}: URL–wikilink conflict (help) – pmc, title, url, title-link
• 1111 (new): "Title". Journal. PMC 12345. {{cite journal}}: URL–wikilink conflict (help) – pmc, title, url, title-link

Presumably you intend to change:

if config.CitationClass == "journal" and not is_set(URL) then

to something akin to:

if not (is_set (TitleLink) or is_set(URL)) then

If that is your change, then the only thing that changes in the examples above is example 1101 which will not show an error. That does not make my wrong as you claim it to be.

You wrote that [I am] trying to instrument this corner case to force the introduction of a parameter to disable auto-linking. Where have I ever said that? In general I am opposed the the introduction of special case parameters of any sort into a template system that is already overburdened with too many parameters. It is highly unlikely that I would now start advocating for such a parameter either openly or sub rosa. I am opposed to special-case anything and gnash my teeth when special cases are unavoidable.

|title=, when linked, is an advertisement to see the actual source that is advertised (my words) is not obviously wrong (your words). The purpose of the RfC was to link |title= to the url created with the |doi= identifier when the source at that doi-identifier-url is free-to-read. A common rationale expressed at the RfC is that a linked title makes it easier for readers to get to that source when they might hesitate to click the doi identifier link because they don't know what a doi identifier is. The linked title is then an advertisement that says to these hesitant readers, "click me, I am a link to the source." Yes, I know about external link icons and the differences between external and internal link colors. That does not change the fact that readers, even those of us who are experienced with how en.wiki citations are rendered, will click blue-linked titles and be disappointed/astonished/frustrated to land at en.wiki's article about the source.

It is, in my opinion, poor practice to link a cs1|2 title to an article about that title. The purpose of a citation is to help readers locate the source that supports content in an en.wiki article. Articles about a source at en.wiki are not WP:RS but linking to such articles through a link at |title= makes it seem that the en.wiki article is the source (after all, it has the blue link). If it is important to mention a particular source in an article, then that mention should be in the body of the article or in an article footnote with standard wikilinks to the en.wiki article about the source. There is a use for |title-link=: citing sources at Wikisource because the link is to the source and the source is reached though standard interwiki links:

{{cite encyclopedia |title=Aard-vark |title-link=s:1911 Encyclopædia Britannica/Aard-vark |encyclopedia=Encyclopædia Britannica |year=1911}}

"Aard-vark" . Encyclopædia Britannica. 1911.

and, no, I don't think that using |title-link= to point to Wikipedia articles should be forbidden. Limited, certainly; forbidden, no.

—Trappist the monk (talk) 23:31, 25 May 2020 (UTC)

Your opinion that titles should not be linked to articles about the title is wrong, and you should feel bad about holding such a wrong opinion. My opinion, which is obviously the correct opinion, is that such links should always be made whenever we have an article about the reference that can be linked from its title. There are very good reasons for linking to an article about a reference: for instance, the article may include multiple free-to-read links where a reference can only have one, or the article may (and probably should) include opinions from the reviewers that may be relevant for how reliable it is as a reference (for an example see Osen's Women in Mathematics, which has sometimes erroneously been used as a reference here). Your opinion is also at odds with the very existence of the title-link parameter. More importantly, it is an editorial decision that should be made by the editors of articles, not enforced by fiat by the behavior of a citation template. If the editors of an article want to control how a title is linked by using a title-link parameter, for a reference that has a pmid or open doi, they should be allowed to control it, just as they can control it by using a url parameter. If the editors of an article for whatever reason decide that a title should not be linked, even though the default would be to auto-link it, they should be allowed to. You should not be inserting your opinions about this into the behavior of the templates. —David Eppstein (talk) 00:04, 26 May 2020 (UTC)

@Trappist the monk: Alright, let me put this differently. No matter whether you understand the problem with your change or not: you have in front of you three editors who disagree with it, and nobody to support it. You should recognize this situation and act accordingly. If you do not, you know what to expect. − Pintoch (talk) 06:28, 26 May 2020 (UTC)

Ok, I take that back, let me just do the change you suggest then - I understood that you were against it. − Pintoch (talk) 06:36, 26 May 2020 (UTC)

Please do not interpret the silence of some of us as agreement with one of the many positions taken above. This discussion appears to have gone far afield from its original purpose, which was to auto-link the title when a DOI was present, unless I am mistaken. It would be great if discussions could stay on topic. If there is a problem with |title-link=, perhaps it should be addressed in a separate thread. – Jonesey95 (talk) 06:51, 26 May 2020 (UTC)

Yes, this was just a corner case of the existing auto-linking mechanism, we should have had this discussion elsewhere in the first place. − Pintoch (talk) 07:00, 26 May 2020 (UTC)

Huh? This is not a corner case. The discussion belongs exactly here. It shows that the whole original proposal, which led to the RfC, was ill-defined. An RfC which cannot be implemented as proposed, because it conflicts with other functionality, is basically useless. What we can still extract from the RfC, even if it was not thought through, is the core message that many people want some form of auto-linking. But we are not going to break existing functionality for this, as the voters in the RfC certainly didn't intend to break existing functionality either. Instead, we will have to find solutions which solve all the issues brought up here before any of this can go live.

--Matthiaspaul (talk) 09:37, 9 June 2020 (UTC)

Auto-linking for book chapters

One issue with auto-linking for book chapters is that the DOI can potentially refer to the entire book or the individual chapter. If no |chapter= is present, I think we can auto-link the title safely. If a chapter is specified then there is a risk that we link the book title with a chapter DOI or that we link the chapter title with a book DOI. The simplest solution I can think of is to disable auto-linking when |chapter= is present, what do you think about that? − Pintoch (talk) 07:00, 26 May 2020 (UTC)

The DOI is supposed to refer to the work actually being cited, so I don't see a problem with linking it. If the citation contains the DOI of the wrong work, it will need to be fixed but there's nothing the template can do to guess it. Nemo 19:15, 7 June 2020 (UTC)

Yeah, but if the DOI only belongs to the |chapter=, not the |title=, the auto-linking should be applied to the chapter rather than the title. Otherwise it would be the same as linking |chapter-url= to |title= even if |chapter= is present. That would be really messy.

--Matthiaspaul (talk) 09:57, 9 June 2020 (UTC)

While this is complicating things even further unfortunately, you bring up a valid point. So far, I thought chapter DOIs would be just some "unofficial" private extension to DOIs by some publishers, but if they are not, we either need some means to declare the type of a given DOI or to provide up to two of them just like we have |url= and |chapter-url=: Perhaps |doi=/|work-doi= and |chapter-doi=?

Disabling auto-linking when |chapter= is present would create even more "unnecessary complexity", making the whole auto-linking thing look unpredictable to normal users. --Matthiaspaul (talk) 09:57, 9 June 2020 (UTC)

A related discussion: Help_talk:Citation_Style_1#Chapter-id_or_Chapter-doi_parameter?

--Matthiaspaul (talk) 11:38, 13 June 2020 (UTC)

Proceeding

I understand that the code has been further refined, so we're ready to go, right? Nemo 19:15, 7 June 2020 (UTC)

I don't think so. We still need some generic means to override and optionally disable the auto-linking behaviour (f.e. by the suggested |title-link=none/identifier_parameter_name/[((]article_name[))] extension), and to sort out what to do with chapter DOIs (f.e. by adding |chapter-doi=).

--Matthiaspaul (talk) 10:11, 9 June 2020 (UTC)

As far as I am concerned this is ready to deploy. There is currently no problem with chapter DOIs since the mechanism is only implemented for {{cite journal}} for now - once this is rolled out we can gradually expand to other citation templates if there is consensus around the course of action for chapter DOIs. − Pintoch (talk) 08:37, 14 June 2020 (UTC)

In the current implementation, there is still no way to (optionally) disable auto-linking or to override the auto-selection of PMCs or DOIs. Also, in other threads people are already asking about adding auto-linking support for S2CID as well. This clearly indicates that we need a proper generic solution instead of continuing to add kludges on kludges only complicating things on all fronts (in the implementation, in the documentation for users, and in the maintenance of citations) further in the long run.

In another thread, I was questioning the very need of existence of the |title-link= parameter, because, as was discussed in that thread, |title= allows for internal links as well (even more flexible than via |title-link= - so, if that parameter would have been removed to reduce unnecessary complexity, my proposal to overload the |title-link=none/<identifier_parameter_name>/[((]internal_link[))] functionality to also control the auto-linking behaviour would have been voided as well, however, I meanwhile have come to the conclusion that |title-link= is still needed to link a combined title if both |title= and |script-title= are used at the same time, therefore my suggestion still remains a possible solution, fortunately. If I understood Trappist correctly, he suggested to use something like |url=none/<identifier_parameter_name>/external_link instead. While the parameter name is less intuitive IMHO, it would even have the advantage that the auto-linking of titles and chapter titles could be controlled individually by something like |chapter-url=none/<identifier_parameter_name>/external_link, whereas in my proposal this would be implicit.

(There are two more cases to be properly addressed in the future: Works without any title at all, and works with a so called descriptive title. At present, {{cite journal}} already implements a special case |title=none to specify a non-existing title. The discussion of these cases is not necessarily related to the question how to control the auto-linking behaviour above, but depending on what solution we would actually go for, it might be possible to address this in one coherent way in order to keep the user interface as easy and intuitive to use as possible. However, at present, my suggestion how to address these two cases would not involve |title-link= etc. at all, but other users might have other ideas.)

--Matthiaspaul (talk) 16:04, 14 June 2020 (UTC)

@Matthiaspaul: I encourage you to run RFCs to get approval for the changes you propose. I suggest we do it only after the current version has been rolled out. − Pintoch (talk) 19:54, 14 June 2020 (UTC)

Following-up on the proposals how to implement the auto-link override, I searched old threads for other proposals and found that Trappist the Monk already suggested |url=none/<identifier_parameter_name>/external_link and Headbomb |autolink=no/yes/<identifier_parameter_name> back in 2016. In 2019, Headbomb suggested |auto-url=none, whereas Nardog and I proposed |url=none/<identifier_parameter_name>/external_link, followed by my newer proposal for |title-link=none/<identifier_parameter_name>/[((]internal_link[))]. These proposals are all very similar in nature. In order not to introduce yet another parameter for this, I think, overloading either |url=/|chapter-url= or |title-link=/(|chapter-link=) is the way to go. Overloading the url parameters could have the disadvantage of temporarily causing trouble for bots trying to make sense of these special values. Overloading |title-link= (without introducing |chapter-link=) leaves open the question of how to best cope with chapter identifiers. Therefore, if the potential bot issue could be ruled out or is found to be minor, I would also support Trappist's proposal of overloading |url=/|chapter-url=. Opinions?

--Matthiaspaul (talk) 14:03, 15 June 2020 (UTC)

Just like I respected the few who opposed my initial proposal on this page and required me to go through an RFC process for a change that turns out (surprise) to be very consensual, please respect my own opposition to your proposal. Is it too much to ask for? − Pintoch (talk) 18:25, 15 June 2020 (UTC)

If you cannot draw this from the very fact that I spent considerable time thinking about a solution how to properly integrate the auto-linking feature although I do not consider it useful, let me state that I do respect your opinion. That doesn't mean that I think it is correct.

However, as your RfC did not define the scope of the auto-linking feature and since it ignored addressing important real-world cases (which, however, need to be solved in an actual implemention), the only thing that can be drawn from the RfC reliably is that many people want some form of auto-linking, leaving it up to us to find a proper solution how to integrate it into the existing citation framework. Specifically, what cannot be drawn from that discussion is that they want to enforce autolinking without any means to override it, as you now seem to push for.

In that discussion, here, and in prior discussions over several years, the whole auto-linking idea was always discussed under the assumption that there is some means to override it where necessary. This was requested even by people who think auto-linking is a good idea. It is obvious that we need this before auto-linking can go live. While other implementation details (such as adding support for S2CID or extending the scope beyond {{cite journal}} and thereby indirectly also addressing the case of chapter identifiers) can be delayed, the implementation of some override functionality can not be delayed any further as we now have two rather only one identifier to distinguish (PMCs and DOIs), and not having any means to override the behaviour when the auto-selection would select and link to the wrong identifier is violating WP:CITEVAR and WP:RS. It would destroy the trust we put into citations which were deliberately set up in a certain way and not in another by those who provided the citations originally.

So, for as long as no means to optionally override auto-linking are implemented, I oppose rolling out the currently half-finished implementation - as is, it is potentially harmful to the project.

Let's solve this issue by thinking through if overloading |url= (or |title-link=) is an extensible solution addressing all potentially desired future cases (I think it is) and not causing problems for bots (not sure about that). And if so, then let's implement it, so that auto-linking can go live without causing harm.

--Matthiaspaul (talk) 19:57, 15 June 2020 (UTC)

I have no interest in continuing this argument. I will just let other editors judge your attitude on their own. − Pintoch (talk) 07:29, 18 June 2020 (UTC)

Screenshot for reference: . Nemo 15:37, 12 July 2020 (UTC)

Where are we on this?

What's the progress/holdup on making all |id-access-free= auto-link? It's been over a month since there was movement on this. Headbomb {t · c · p · b} 17:29, 4 August 2020 (UTC)

trans-title= splits PDF format indication

In the following example, the "(PDF)" format designation comes after the translated title, which looks odd because the PDF symbol is displayed after the foreign language title:

Beskorovainyi, Vladimir V.; Soboleva, Elena V. (2010). ИДЕНТИФИКАЦИЯ ЧАСТНОй ПОлЕЗНОСТИ МНОГОФАКТОРНЫХ АлЬТЕРНАТИВ С ПОМОЩЬЮ S-ОБРАЗНЫХ ФУНКЦИй [Identification of utility functions in multi-objective choice modelling by using S-shaped functions] (PDF). БИОНИКА ИНТЕЛЛЕКТА [Bionics of Intelligence] (in Russian). Vol. 72, no. 1. Kharkiv National University of Radioelectronics. pp. 50–54. ISSN 0555-2656.

There are several potential ways to solve this:

1. Include the translation in the link - undesirable because the translation does not actually belong into there.
2. Move the (format) designator in front of the translation - undesirable because it looks out of place there
3. Move the PDF symbol after the translation - undesirable because it also serves as "external link" indicator, also not sure if this is technically possible
4. Suppress the PDF symbol and replace it by the normal "external link" symbol. Do we really need the PDF symbol, anyway? (Do we support any other file types with special symbols?)

Ideas? --Matthiaspaul (talk) 14:28, 20 June 2020 (UTC)

#4 is possible. Yes, other file types do have special symbols if the URL is a certain way and file type. PDF is understandably the most common. Because I have the link handy for Modern skin, the stylings applied to the others are in main.css. The other skins have similar rules, if not exactly the same. (See also phab:T225430.)

#3 is technically possibly. It would require including the trans-title in the link. (See also prior discussion.) If we do not want a sea of blue on this route, that is also technically possible but it would require an inaccessible design choice (to wit, making the link for some part of the title + trans-title not blue).

I would favor #2 if there is a consensus that this needs to change. It's puzzling that you think this undesirable given our recent discussions on parentheses stacking... ;) --Izno (talk) 14:59, 20 June 2020 (UTC)

Thanks for the links, Izno.

Well, I don't like #2, because it separates the actual title and the translation too much. To me, they belong together logically and therefore should not be separated - just like the format symbol and "(format)" text should be displayed in one location. To me, title information is far more important than format information, therefore, format information should not be inserted in the middle of title information, and in particular not given in two places.

Therefore, of the given choices, I would prefer #1 (which implies #3), that is, back to the old 2017 behaviour. Although I am not particularly fond of the potentially resulting long blue link labels, they would be of only minor concern to me (display cosmetics) - at least, they are logical.

For completeness, another solution would be to swap title and trans-title, but I don't actually suggest this, because I think the actual title should be displayed before the translation.

--Matthiaspaul (talk) 16:21, 20 June 2020 (UTC)

@Matthiaspaul: I've corrected the citation formatting above (but not in a way that affects the topic of this discussion). I hope you don't mind. Glades12 (talk) 07:03, 13 July 2020 (UTC)

Good to know that we meanwhile support |script-*= and |trans-*= parameter variants for them. I've changed it to |magazine=, though - they describe themselves as a magazine rather than as a journal. |work= is too unspecific. I use |work= only when none of the more descriptive parameters applies (typically with {{cite web}} or {{cite book}}, rarely with {{cite journal}} or {{cite magazine}}). --Matthiaspaul (talk) 10:22, 13 July 2020 (UTC)

There would be yet another way how this could be solved: Given that providing the PDF icon and a "(PDF)" text is redundant, we could simply suppress the "(format)" text if it resembles one of the external link types recognized and indicated by specific icons. If |format= is used to specify something different, it would be displayed as before, but then the icon and the text would not be redundant and therefore look much less out of place than now. --Matthiaspaul (talk) 18:12, 20 July 2020 (UTC)

It is not a given that the PDF icon and a "(PDF)" text is redundant. The pdf icon is rendered by MediaWiki as a css background-image property (MediaWiki:Common.css). Because it is not rendered from an html <img>...</img> tag, it does not support the alt= attribute. In cs1|2, the automatic |format=PDF is a way to notify screen-reader-users that the source is a pdf file. That functionality should not be degraded.

—Trappist the monk (talk) 18:56, 20 July 2020 (UTC)

I agree, accessibility is important. Still, there are potential remedies for this as well:

Instead of completely muting the "(format)" text when it would be redundant because of the icon, the text could then be made invisible to normal users, but left readable for screen-readers only:

• https://webaim.org/techniques/css/invisiblecontent/
• https://webaim.org/techniques/alttext/#background

Or, the global link decoration could be disabled and the PDF icon be displayed locally (as image, and with alt= text).

--Matthiaspaul (talk) 22:57, 20 July 2020 (UTC)

Proposed change to |script-xx= parameters

I propose that the namespaces of the parameters |script-title=, |script-chapter= and |script-work= be changed to use ISO 15924 codes instead. This has the benefits of eliminating ambiguity (for instance, whether a Chinese-language source is written in traditional or simplified characters, or whether a Mongolian one uses the Cyrillic or Traditional Mongolian writing system) as well as making it easier to properly display obscure scripts which are not presently known by CS1, but coded by the ISO. (See Category:Indonesian scripts for starter examples of the latter.) Glades12 (talk) 14:52, 22 June 2020 (UTC), updated 14:52, 1 July 2020 (UTC)

The HTML standard requires BCP 47 (see [4]). --Izno (talk) 15:14, 22 June 2020 (UTC)

Why? The scripts are what may be rendered wrong, not the languages themselves, right? Or am I completely mistaken? Glades12 (talk) 19:12, 22 June 2020 (UTC)

You are not mistaken. The proposal seems beneficial and uncontroversial. 65.88.88.69 (talk) 13:32, 23 June 2020 (UTC)

@Izno: You still haven't explained. Glades12 (talk) 17:55, 5 July 2020 (UTC)

Please read the provided links. If you still have questions, please let me know. --Izno (talk) 18:49, 5 July 2020 (UTC)

I have read both of them several times, and still do not understand how either supports the current system. Is the problem that the title needs to be pronounced correctly or accommodate users of different fonts? If that was the case, we would logically also require specific language codes in |title=, |quote=, etc.. The fact that we don't leads to the conclusion that only the scripts themselves are necessary to specify. Glades12 (talk) 19:36, 5 July 2020 (UTC)

The specification requires BCP 47. The syntax we output and accordingly require as input should conform to BCP 47 accordingly. Is that fact confusing? How so? --Izno (talk) 20:29, 5 July 2020 (UTC)

Template:MongolUnicode seems to work pretty well without it. Glades12 (talk) 06:55, 6 July 2020 (UTC)

That comment does not answer my question. --Izno (talk) 11:56, 6 July 2020 (UTC)

I don't know about you, but to me it shows that ISO 639 codes are not absolutely necessary just to display a script correctly. Glades12 (talk) 07:28, 7 July 2020 (UTC)

I did not argue that ISO 639 codes are necessary. Nor did you, until just now, indicate what you were attempting to do. I have no objection to BCP 47 codes, but the overhead associated with MongolUnicode is frankly unnecessary for this template for all languages and I would oppose on that basis an implementation of CSS in these templates. I suspect if you had actually comprehended BCP 47 you would have observed that ISO 15924 is included as valid in the BCP, so the request to add accepting those codes seems valid, but you somehow have not brought that up yet. --Izno (talk) 16:19, 7 July 2020 (UTC)

I'll be happy to read ISO 15924 just as $61.63 in the form of United States currency or a United States postal money order arrives in my physical mailbox. Until them I'm opposed. Jc3s5h (talk) 16:09, 7 July 2020 (UTC)

There is literally no way to know that from your links (do you expect everyone you meet to be a HTML genius?), and the way you replied made it seem like you did disagree. Anyway, sorry for the misunderstanding. Glades12 (talk) 12:47, 10 July 2020 (UTC)

Don't add a doi.org link if you're using the |doi= parameter

To our citation template documentation such as Cite journal/doc#URL, could we add a sentence explaining that one should not add |url=http://dx.doi.org/10.2307/1440636 if there's already |doi=10.2307/1440636 defined, as was for instance here? Same goes for JSTOR or Worldcat (OCLC), where we don't need |url=http://worldcat.org/oclc/777999581 if there's already |oclc=777999581, as here. --bender235 (talk) 17:47, 24 June 2020 (UTC)

That doesn't seem like a common enough error to warrant yet another addition to a group of pages that are already so bloated with instructions that they're barely navigable. We should focus on cutting the CS documentation down. Glades12 (talk) 10:46, 25 June 2020 (UTC)

Depends on your definition of "common." I've been fixing this issue over the past couple of days with AWB. The DOI-in-URL issue alone affects about 2,000 articles, and I was just scanning for the outdated dx.doi.org scheme. --bender235 (talk) 16:41, 25 June 2020 (UTC)

There may be issues here. See § Auto-linking titles with free DOIs and the related RfC. 64.61.73.83 (talk) 20:12, 25 June 2020 (UTC)

Could you please elaborate on what issues you expect? The way I understand the linked RfC (which I missed, unfortunately, but in retrospect wholeheartedly support) a link is being placed under the article title if |doi= and |doi-access=free is set, but it does not mean we actually need to place anything in the |url= parameter. In fact, anything put in there would overwrite the DOI link. --bender235 (talk) 16:52, 28 June 2020 (UTC)

Mainly that the AWB edits may be at cross-purposes with the new auto-linking feature, or perhaps redundant. I see similar/additional concerns are discussed directly below. Not disputing the merit of your proposal, however. 98.0.246.242 (talk) 20:59, 29 June 2020 (UTC)

Okay, maybe it's enough then. Glades12 (talk) 07:19, 26 June 2020 (UTC)

This only holds true if the url added would be exactly the same as the link provided by doi. For example, if

|doi=10.1145/360569.360660

is already present in a citation, it does not make any sense to add:

|url=https://doi.org/10.1145%2F360569.360660

In many cases, it also would not make sense to add

|url=https://dl.acm.org/doi/10.1145/360569.360660

(the link the doi resolves forwards to at present), because the forwarding might go down in the future or resolve to something else, so the links are not exactly redundant - however, that's debatable.

However, even with |doi=10.1145/360569.360660 being present in a citation (and with DOI auto-linking enabled in the future), it still makes a lot of sense to add |url= if it points to a better source or even the actual document rather than only a metapage, so

|url=https://dl.acm.org/doi/pdf/10.1145/360569.360660

(a very similar looking link, but pointing to the PDF, rather than only to the metapage) is still appropriate to add, in particular since

|archive-url=https://web.archive.org/web/20200624113752/https://dl.acm.org/doi/pdf/10.1145/360569.360660

can be added as well then (it cannot for |doi= alone).

Example without url (in the future with DOI auto-linking):

Chen, Tien Chi; Ho, Irving Tze (January 1975). "Storage-Efficient Representation of Decimal Data". Communications of the ACM. 18 (1): 49–52. doi:10.1145/360569.360660. S2CID 14301378.

Preferred example with url:

Chen, Tien Chi; Ho, Irving Tze (January 1975). "Storage-Efficient Representation of Decimal Data". Communications of the ACM. 18 (1): 49–52. doi:10.1145/360569.360660. S2CID 14301378. Archived from the original on 2020-06-24. Retrieved 2020-06-24.

--Matthiaspaul (talk) 23:39, 25 June 2020 (UTC)

@Matthiaspaul: I was in fact referring to URLs that are exactly what the DOI (or OCLC, or PMID, etc.) parameter creates. --bender235 (talk) 16:52, 28 June 2020 (UTC)

Yeah, there's consensus to remove (or not add) exact matches, however, some people incorrectly assume this would also apply to "similar" links. --Matthiaspaul (talk) 08:13, 7 July 2020 (UTC)

This is already covered in Template:Citation Style documentation/id2 per long-standing consensus. Where else would you like to write it? Nemo 06:31, 13 July 2020 (UTC)

Some maintenance items to upgrade to errors

Right now, Category:CS1 maint: display-authors and other friends are nearly always empty because they are nearly always an easy-to-correct error. I would like to propose upgrading them to errors accordingly, which will make them more visible to editors.

To make this easier to do in the future with maintenance messages we decide should be errors, I'd also like to see the error and maintenance system implementations be made the same (save for the obvious distinction). For this latter, I trip up really hard every time I want to get maintenance items turned into errors, and it's making it hard to parse how to make the necessary code modification for display-X.

Any concerns? --Izno (talk) 12:45, 4 July 2020 (UTC)

After the next update which I am about to announce.

Error messages rely on tables defined in Module:Citation/CS1/Configuration. Each error message has four properties: message, anchor, category, hidden. For maint cats, we might define these:

message always nil or empty string

anchor a unique value used to link into Help:CS1 errors help text

category the key that tells common set_error() how to handle the issue; maint cat names all begin with 'CS1 maint:' whereas error message cats aren't so consistent

hidden ignored for maint cats which are always hidden

Category:Pages with inactive DOIs is presently a member of Category:CS1 maintenance but doesn't necessarily belong there. In Module:Citation/CS1/Identifiers this cat and its dependents are treated as pseudo errors (the categories are created and added to the table z.error_categories. Property and maint cats have their own tables (maintenance_cats and properties_cats). Maint cats can display their names as an editor-option via css, prop cats display nothing.

Before we embark on a messaging rewrite, we should normalize, somehow, the inactive doi handling. Inactive dois are not errors in the sense of cs1 errors like any of the categorized errors in Category:CS1 errors. Neither are they simple properties because they should be fixed. We could treat inactive dois as maint issues so that editors who have turned on maint messaging can see locate the inactive dois.

Another thing that I would like to do is standardize the location of error messages. We have a mix of locations: some error messages are adjacent to where the element occurs in the rendered citation and the other are all listed at the end. I have a preference for grouping all of the error messages at the end. Maint messaging is all rendered at the end of the citation.

Assuming that we do all of this, the simple case conversion from maint message to error message is to change category to the correct error cat; add the appropriate error message in message, and set hidden to the appropriate boolean.

No doubt, this enumeration is incomplete ...

—Trappist the monk (talk) 15:30, 4 July 2020 (UTC)

Speaking of error messages, it might be worthwhile to also improve the preview messages on duplicated parameters so that they provide some location information. In articles with many citations it sometimes takes quite a while to spot the citation using f.e. |date= twice. --Matthiaspaul (talk) 07:58, 7 July 2020 (UTC)

@Matthiaspaul: That's a MediaWiki feature, unrelated to the citation templates. But there's a script, User:Frietjes/findargdups, which can tell you which template call has the duplicate. -- John of Reading (talk) 08:42, 7 July 2020 (UTC)

Thanks, John. --Matthiaspaul (talk) 14:27, 7 July 2020 (UTC)

The error message function has been renamed set_error() → set_message() and been modified so that it will emit maintenance category 'messages' when the message property for that message is nil. Maint messaging is now part of the error_conditions{} table in Module:Citation/CS1/Configuration/sandbox; TODO: rename that table.

Module talk:Citation/CS1/testcases2

—Trappist the monk (talk) 14:37, 30 July 2020 (UTC)

I worked my originating request as a result of this rework, which moves display_names messaging (inconsistently error and maintenance) to strictly errors. Main, /Configuration and testcases2. Perhaps of note that this changes the error for the "don't know" case from invalid_param_values to disp_name; I made that change to centralize all the disp_name errors fixing. There may be further work that should be done so the other use of that error can be more definite. --Izno (talk) 22:50, 8 August 2020 (UTC)

One other thing: I am not personally convinced that we need 5 categories to handle display_names issues. I think 1 would suffice. Seeking feedback on that point. --Izno (talk) 22:52, 8 August 2020 (UTC)

module suite update 11–12 July 2020

I propose to update cs1|2 module suite over the weekend 11–12 July 2020. Here are the changes:

Module:Citation/CS1:

• Vancouver Style non-Latin character error message fix; discussion
• add cs1 and cs2 class attributes; discussion TODO: currently disabled because breaks every test at Module talk:Citation/CS1/testcases; restore before next update
• remove separate contribution alias support; discussion
• separate encyclopedia parameter aliases from periodical aliases; discussion
• remove separate section alias support; discussion
• provide default value for |url= if |doi= and |doi-access=free are present; discussion and discussion
• tweak {{cite conference}} error messaging; discussion

Module:Citation/CS1/Configuration:

• remove separate contribution alias support
• separate encyclopedia parameter aliases from periodical aliases
• moved identifier limits into handler tables; discussion
• remove separate section alias support
• tweak trans-missing-title error messaging; discussion
• add limited support for quarterly dates; discussion and continued
• add Easter as a named date; discussion (linked discussion)

Module:Citation/CS1/Whitelist

• template-specific args table for {{cite map}}; discussion
• add |trans-title= to limited param list; discussion

Module:Citation/CS1/Date validation

• fixed disambiguated-date reformat bug; discussion
• add limited support for quarterly dates

Module:Citation/CS1/Identifiers

• moved identifier limits into handler tables
• ISMN label to use redirect; discussion
• wikidata code optimization; discussion
• increase doi-registrant limit; discussion

Module:Citation/CS1/COinS

• add limited support for quarterly dates

—Trappist the monk (talk) 14:10, 5 July 2020 (UTC)

• And a few tweaks to parameter suggesting from Module:Citation/CS1/Suggestions/sandbox. --Matthiaspaul (talk) 09:17, 7 July 2020 (UTC)
• For the class attribute, do we also want to have them populate Category:Pages using CS1/Category:Pages using CS2 or something, so we can do category intersections to easily find pages that use a mix of styles? Headbomb {t · c · p · b} 13:39, 7 July 2020 (UTC)

You can more or less already do this with special:search with a little guess and check e.g. citation and cite book. --Izno (talk) 14:04, 7 July 2020 (UTC)

Which won't detect those with {{citation |mode=cs1}} or {{cite book |mode=cs2}} properly. The categories would. Headbomb {t · c · p · b} 14:19, 7 July 2020 (UTC)

So the cs2 cat would have roughly 300k pages (all current {{citation}} templates + some hand-waved-number of cs1 templates with |mode=cs2) and the cs1 cat would have 4500k - 300k = 4200k pages? I'm not really seeing the benefit of that.

—Trappist the monk (talk) 14:38, 7 July 2020 (UTC)

Even if there's benefit, there is a significant cost to most pages where we would now be including a category in every citation when output. --Izno (talk) 14:56, 7 July 2020 (UTC)

So? It's just another hidden category. If you don't like it, ignore it. Because the benefits would be tangible, and we'd have a way to keep track of articles with a mix of CS1 and CS2. Headbomb {t · c · p · b} 16:23, 7 July 2020 (UTC)

I'm not talking about the visible cost, I'm talking about using a template to include, in an article with 300 citations, 300 x category links (that the skin will process to 1 in its output but which are still counted against the page in the template transclusion cost). Regardless, at this point I think this request is reasonably outside the scope of this update. If you would like to pursue further discussion, please feel free to break this into a new section like I did below. --Izno (talk) 16:27, 7 July 2020 (UTC)

• In regard to the requested auto-linking feature (Proceeding), while I understand that some people want to see this being rolled out as soon as possible (to the extent of suppressing concerns and making snarky remarks), I do consider the current implementation half-baked as it does not provide any means to override the automatic behaviour in cases where this would become necessary. As even supporters of the feature suggested some possible override (in the current discussions, but also in discussions going back to 2016), I consider it important to implement this before rolling out the feature. Using Trappist's original parameter suggestion for this as an example (which I consider to be even more future-proof than my own proposal due to the postponed |chapter-url= issue), what would be needed as a minimum now is to let |url= take three additional values: none, doi and pmc. doi and pmc would select the corresponding identifier link even if auto-linking based on its "priority ruleset" would select a different identifier, and none would disable auto-linking for this citation. For other url values, the parameter's argument would be used as a link target. And without |url= at all, auto-linking would work according to the ruleset discussed (which has already been implemented for normal titles in journals the least).

Question is if someone could still implement this reliably before the weekend, if the rollout should better be delayed a week or two until this has been implemented, or if this should be rolled out without override facility for now?

--Matthiaspaul (talk) 15:06, 7 July 2020 (UTC)

|auto-url=doi/pmc/none would be great. Supporting all the free identifiers (of record) would be even better. But I wouldn't delay the update just for that, with the understanding that this templates should be updated way more often than it currently is for major features like this. Headbomb {t · c · p · b} 16:39, 7 July 2020 (UTC)

Yes, for a coherent interface the idea is to be able to address other identifiers as well in a consistent way in the future, although IMO only with "manual" auto-linking instead of through some arbitrary automatic rule-set.

You really suggest a new parameter |auto-link= for this? So far, I took your parameter name as a quick "discussion handle" or "prototypical name", only...

I was suggesting to overload |title-link= with this functionality in order to avoid introducing yet another parameter, and Trappist originally proposed to overload |url= instead. Could you live with that as well? The basic idea behind that is to make existing parameters more functional in a "smart", coherent and easy-to-remember way rather than to introduce new narrow-purpose parameters. Overloading existing parameters is a bit more difficult to code than using new parameters, but I think what is more important is the user-interface side - what is more intuitive to use and easier to document?

While using |title-link= (or |auto-url=) rules out any possible problems with bots needing an update after rolling out this feature, |url= has the advantage that (at a later point in time) the feature could be extended to auto-link chapters (this was already requested in the discussions above) by overloading the already existing |chapter-url= parameter in the same way we'd do it for |url= (whilst we don't have nor need an equivalent |chapter-link= parameter, and given that title and chapter could point to different identifier links (f.e. book and chapter DOIs) it would be difficult to control this behaviour with a single parameter like |auto-url= only).

--Matthiaspaul (talk) 15:49, 8 July 2020 (UTC)

Not fussy on the actual name of the parameter, nor do I see big downsides to using the existing one. And automatic linking should be done whenever possible (and that it makes sense to do so). If there are multiple free identifiers of records just have a default order, which can be overiden by |auto-url=/|whatever=. Headbomb {t · c · p · b} 22:54, 8 July 2020 (UTC)

Thanks for the clarification of your position. The default order is exactly what I have issues with (in particular if more than two identifiers will be involved), that's why I am in favour of "manual" auto-linking at least any further supported identifiers. What might appear as an obvious and convenient order of priorities to some might appear as completely non-sensical and counter-productive to others. However, what is most important is that it will be possible to override any automatic behaviour where necessary. Good that we agree on this. It will help to implement this feature in a way which is acceptable for anyone.

--Matthiaspaul (talk) 13:23, 9 July 2020 (UTC)

If you have issues with the 'default' order, then override it for whatever article the default order is not ideal. That we're linking via the DOI or via the PMC is really inconsequential to the reader, which will have access to the same full free version of record either way. Headbomb {t · c · p · b} 13:31, 9 July 2020 (UTC)

Yeah, that's fine and that's what I will do if I run into the situation (and there will be an override facility).

However, my point was that I think an automatic rule-set doesn't scale. It might still make a reasonably good selection for PMCs and DOIs, but with more auto-link identifiers added to the list in the future there will soon be a point where no reasonable priority order can be found any more for a generic audience, it will just be arbitrary and appear as pseudo-random to people. Consequently, there will be a growing number of people who need to override the automatic order, to the point that it does not make sense to auto-select one identifier in the first place, but just let editors manually select the desired one.

Either way, that's future stuff, what we need now is to consolidate the current implementation.

--Matthiaspaul (talk) 12:49, 13 July 2020 (UTC)

Editors

Rats, I thought I might be able to make deprecating |editors= in this release. Just south of 800 to go. :^) --Izno (talk) 02:50, 7 July 2020 (UTC)

I'd be against deprecating |editors= much like I'd be against deprecating |authors=. They're convenient when you don't have the time or willpower to add |editor-last1=/|editor-first1= to |editor-last#=/|editor-last#= to a long list of editors. Headbomb {t · c · p · b} 13:36, 7 July 2020 (UTC)

It is already deprecated by the presence of Category:CS1 maint: uses editors parameter. --Izno (talk) 14:03, 7 July 2020 (UTC)

That's not deprecation, that's marking the usage as "not ideal, could probably be better". Very different things. Headbomb {t · c · p · b} 14:35, 7 July 2020 (UTC)

I have yet to be reverted in some 4000 edits removing the parameter. That's functionally deprecated, even if "CS1 maint" didn't give a strong clue as to our opinion on the condition existing. --Izno (talk) 14:56, 7 July 2020 (UTC)

I'd be against deprecating those two parameters also. (And by the sound of it, they're not deprecated, strictly speaking, just advised against.) Authors= and Editors= allows for situations where there are multiple individuals to name but some are credited under "with", ie they appear to be afforded a lesser credit (and a smaller font size) on a book cover and title page. This is often the case with autobiographies (for journalist co-writers) and revised editions of multiple-author works (where a previous editor might be credited under "with" because content from a previous edition is reproduced without any change). Besides, as ever, I wish template editors here would stop trying to jam an overly simplified, one-size-fits-all model down everyone's throats ... The important thing is to present the source accurately. JG66 (talk) 15:01, 7 July 2020 (UTC)

I would be shocked if you need "with" to do so. Even if "with" is important, and you don't want to include that person as an |editorn=, you have |others= to express in total free-text the full relationship of any other people involved in the work. I have removed several "withs" in my run to remove |editors= ("with", er, replacement as a "full" editor) and again, no reversions if in fact anyone cares that deeply. --Izno (talk) 15:26, 7 July 2020 (UTC)

Well, you'll just have to be shocked then, won't you. I'd happily revert every single instance where including "with" ensures the 'correct' author/editor credit appears. But that's because I write articles rather than gnome away. JG66 (talk) 15:37, 7 July 2020 (UTC)

Including attribution text in the rendering of a citation is one of the purposes of the |<name-list>-mask= parameters – this particular use in mentioned in the template documentation. This citation can be rewritten to avoid |editors= and maintain the proper metadata:

{{cite book/new |last=Marcus |first=Greil |authorlink=Greil Marcus |chapter=The Beatles |chapter-url=https://greilmarcus.net/2014/07/11/the-beatles-1979/ |editor=DeCurtis, Anthony |editor2=Henke, James |editor3=George-Warren, Holly |editor-mask3=with George-Warren, Holly; |editor4=Miller, Jim |title=The Rolling Stone Illustrated History of Rock & Roll: The Definitive History of the Most Important Artists and Their Music |url=https://books.google.com.au/books/about/The_Rolling_Stone_Illustrated_History_of.html?id=ubWAht7N7zsC&redir_esc=y |year=1992 |orig-year=1979 |publisher=Straight Arrow |location=New York |isbn=0-679-73728-6 |via=greilmarcus.net}}

Marcus, Greil (1992) [1979]. "The Beatles". In DeCurtis, Anthony; Henke, James; with George-Warren, Holly; Miller, Jim (eds.). The Rolling Stone Illustrated History of Rock & Roll: The Definitive History of the Most Important Artists and Their Music. New York: Straight Arrow. ISBN 0-679-73728-6 – via greilmarcus.net.

—Trappist the monk (talk) 13:03, 8 July 2020 (UTC)

While I hate running into them in citations, I think, Headbomb has a good argument here for us multi-taskers. Sometimes, they really are convenient: When you are working on something different and don't want to or can't spend time on unrelated stuff, but just want to rush out a citation rather than to not provide the citation at all. However, someone will have to clean up the mess sooner or later. Putting them in a maintenance category is good, but perhaps we could also display a warning in article preview so that whoever edits the article will be made aware of the messy citation. Also, the parameters should be further marginalized in the documentation. This way, we could, perhaps, keep them longtime, but still reduce their use to a minimum.

--Matthiaspaul (talk) 15:02, 7 July 2020 (UTC)

I empathize with the "convenient" argument, but after clearing out 5000 pages (some credit to Ttm who picked up a few hundred), I much-more-deeply empathize with the "as an author, do your job and indicate clearly who is an editor and who is not in the CS1/2 style". :) --Izno (talk) 15:26, 7 July 2020 (UTC)

I'd be down for making most maintenance messages visible in previews (something like Preview message: consider replacing |editors= with |editorn-last=/|editorn-last=). Headbomb {t · c · p · b} 16:28, 7 July 2020 (UTC)

Which is why we have Module:Citation/CS1/Suggestions. --Izno (talk) 16:31, 7 July 2020 (UTC)

Sounds like a good solution. Glades12 (talk) 16:39, 7 July 2020 (UTC)

And just now finished the category off save for some drafts that would be better off G13d. --Izno (talk) 00:28, 8 July 2020 (UTC)

Call me ignorant, but what does G13d stand for?

--Matthiaspaul (talk) 12:45, 21 July 2020 (UTC)

Wikipedia:Criteria for speedy deletion § G13. Abandoned Drafts and Articles for creation submissions

—Trappist the monk (talk) 13:20, 21 July 2020 (UTC)

Thanks. :-) I was somehow trying to match that with numeronyms like I18n or L10n... ;->

--Matthiaspaul (talk) 13:38, 21 July 2020 (UTC)

So we should mark |editors= as deprecated. Because cs1|2 will never be smart enough to correctly parse-apart a string of human names in all of their possible forms, separated by an often inconsistent variety of separator characters, names listed in |authors= and |editors= have never and will never be included in the citation's metadata. It is time to deprecate and remove |editors=. Now seems as good a time as any.

—Trappist the monk (talk) 13:03, 8 July 2020 (UTC)

That will just lead to further abuse of |editor= to contain multiple editors, without the benefit of the maintenance category. Headbomb {t · c · p · b} 13:30, 8 July 2020 (UTC)

Possibly, but such misuse is caught and categorized in Category:CS1 maint: multiple names: editors list (109).

—Trappist the monk (talk) 14:07, 8 July 2020 (UTC)

Does it suppress metadata for as long as this condition applies?

Can this condition be indicated in preview as well?

--Matthiaspaul (talk) 14:21, 8 July 2020 (UTC)

Multiple names in single-name parameters are all shoehorned into a single key/value pair in the metadata because cs1|2 expects, and the documentation requires, one name per parameter. The whole list of names is present but the metadata are corrupted because each name should be in its own k/v pair. As I said before, cs1|2 is not smart enough to parse-apart a list of human names.

If editors have maintenance messaging turned on, the messages are always visible.

—Trappist the monk (talk) 14:47, 8 July 2020 (UTC)

Okay, but if the template can detect that there seems to be more than one name given in a single |author=/|editor= parameter in order to put it into a maintenance category, this could also be used to suppress metadata creation until someone has checked the issue and either splitted the list into many parameters or used (()) to accept the string as valid. With (()), metadata creation would be turned on again, warnings disabled, and the citation would no longer be put into a maintenance catagory.

With this metadata suppression in place and with warnings shown to anyone (not only those who opted in) in preview, I think, we could safely deprecate the |authors= and |editors= parameters. Headbomb's concerns, that people will then use |author=/|editor= to "park" name lists, will likely hold true, but with metadata suppressed and warnings in place, no harm would be done by this and the situation would likely be fixed earlier than having dedicated parameters to take multiple names. Appears like a good compromise and "best of both worlds" approach to me.

--Matthiaspaul (talk) 13:54, 9 July 2020 (UTC)

There are sufficient other maintenance/error categories and correct uses of the templates that I have made changes to the sandboxes to mark editors as deprecated:

--Izno (talk) 14:02, 16 July 2020 (UTC)

Wikidata identifiers

I would like to see a |wikidata= parameter added to CS1. With the advent of things like Scholia (Q45340488) and Template:Cite Q (Q22321052) we now have a large body of scholarly articles indexed in Wikidata. I have manually added a few via |id= (which can be found via Special:WhatLinksHere/WDQ (identifier)). I can understand the resistance to added any and all such identifiers (e.g., Google Scholar paper ID (P4028), Semantic Scholar paper ID (P4011), Microsoft Academic ID (P6366), NII article ID (P2409), etc.; many more can be found at d:Template:Bibliographic properties) however, I believe we should at least support our own WMF identifiers and then the rest can be linked from there. Thank you, —Uzume (talk) 17:20, 8 July 2020 (UTC)

I disagree. The purpose of a citation is to help the reader find a publication. The citation is not meant to be a list of all the places that have cataloged the publication. So please explain how your addition will help the reader find the publication. Jc3s5h (talk) 17:27, 8 July 2020 (UTC)

@Jc3s5h: It sounds like you are advocating more for adding things like Internet Archive ID (P724). Have you looked at some the Wikidata records? For example, Semantics of context-free languages (Q56672530) has many links to actual PDFs of the article via full work available at URL (P953). I do not see how this is much different from the likes of |citeseerx=, etc. —Uzume (talk) 17:31, 8 July 2020 (UTC)

This reminds me of {{cite doi}} and its friends, which were rejected in multiple discussions: Cite doi template substing (Sep 2014) and deprecation of Cite doi templates (Sep 2015). Those discussions mentioned Wikidata as a possible repository for citation information, but I haven't seen any move in that direction. I think the original idea was that something like {{Cite Q}} would be used in place of {{cite journal}} rather than adding wikidata options to CS1 templates. I could be misremembering, though. – Jonesey95 (talk) 17:54, 8 July 2020 (UTC)

@Jonesey95: if it isn't already, {{Cite Q}} could just be a wrapper for CS1 pulling data from Wikidata and filling in fields. —Uzume (talk) 18:02, 8 July 2020 (UTC)

{{cite Q}} uses {{citation}} so that it can do both periodica and book type citations.

—Trappist the monk (talk) 20:03, 8 July 2020 (UTC)

Adding an identifier (as suggested here) is somewhat different than {{cite Q}}, which is the parallel to {{cite doi}}. --Izno (talk) 18:00, 8 July 2020 (UTC)

I'm not enthusiastic about this. There is ongoing discussion (thankfully, mostly elsewhere) where one camp apparently believes that identifiers are wholly ignored by readers because the readers don't know what the identifiers are so won't click them to get to a copy of the source. If that is true, you can be sure that no one will ever follow a wikidata identifier link and, even were they to do so, the wikidata presentation is so user-unfriendly that readers will flee from it.

There is a possible issue of copyright. I notice that the example of Knuth's "Semantics of context-free languages" is behind a paywall at the publisher (doi:10.1007/BF01692511) yet there are 7 purportedly 'free' copies available under the 'full work available at URL' heading at Q56672530. What is the copyright status of all of those?

And then there is the perennial issue of wikidata reliability ...

—Trappist the monk (talk) 20:03, 8 July 2020 (UTC)

Six of the seven copies to which you refer are on .edu domains. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 21:17, 8 July 2020 (UTC)

Just because a University uses a reference in a course and uploads it to the internet for use of its students (presumably under fair-use terms) does not mean the original copyright holder has surrendered their rights. We need to be very careful about linking to such things. What is appropriate fair use for a student on a course is not necessarily fair use for random members of the public.Nigel Ish (talk) 21:38, 8 July 2020 (UTC)

I checked one of the two instances of WDQ (identifier) being used in a CS1 citation, in this case at Virtuality (software design). When I click through to the Wikidata page, I am met with a page whose title does not match that of the cited journal article, and also with a list of seven links to download the article. This article is not freely available; access to its contents is clearly prohibited by the ACM's terms. Such linking is forbidden by Wikipedia policy, which reminds us that Knowingly and intentionally directing others to a site that violates copyright has been considered a form of contributory infringement in the United States. I did not see a corresponding page at Wikidata; clicking on "Wikidata item" from WP's copyright page leads to an irrelevant page, and I did not find a copyright policy linked at d:Wikidata:List of policies and guidelines. I could just be bad at finding things, though.

TL;DR: It appears that this Wikidata entry, chosen at random, is violating US copyright law, as least as described at en.WP's copyright page. I don't see how we can in good faith link to such Wikidata entries, over which we have no control or oversight. – Jonesey95 (talk) 18:36, 9 July 2020 (UTC)

Apparently, according to the discussion at d:Property talk:P953, the "full work available at URL" property under which those urls were listed was, until 2017, primarily intended and documented to be used for free links to non-free works. I.e., piracy. The property documentation now says it should be for "legal online providers" of the resource, but obviously from your example that's not very carefully enforced. —David Eppstein (talk) 19:01, 9 July 2020 (UTC)

It is one thing to criticize Wikidata for copyright issues (linking or otherwise) but I do not see this as substantially different than linking to "S2CID 2556127" which the same reference on Virtuality (software design) also refers to with |s2cid=2556127. You will notice Semantic Scholar also has a links to the full content—possibly in copyright violations as well. How is |s2cid= which already exists better than a potential |wdqid=, |qid=, or |wikidata=? If anything Wikidata is more readily accessibly and such issues could be more easily rectified. If not linking to something because it has issues is a compelling argument, I would advocate we should not allow linking to some other non-English Wikipedias (or even some local articles which are not in good shape in such regards). —Uzume (talk) 04:56, 17 July 2020 (UTC)

I agree that linking to this paper's s2cid page is not substantially different from linking to its Wikidata page. Both the s2cid page and the Wikidata page clearly link to freely downloadable copies of copyrighted material, in apparent contravention of copyright law. I think that we at Wikipedia should not link to that s2cid page or to the Wikidata page from any articles until the links to copyvio material have been removed. – Jonesey95 (talk) 05:40, 17 July 2020 (UTC)

I see your point, but I think you put too much weight on one possibly weak spot completely ignoring the potential benefits.

First of all, most of the listed links on these sites probably go to works which are legal to share, and second, at least in my judgement, links to copies of a work are the least interesting thing why we would want to link to Wikidata specifically:

Wikidata is about semantic organization of data and advanced data retrieval and research methods around this concept - something that is not in the scope of citations at Wikipedia, but nevertheless is in the scope of at least some among our audience. For example: When I read about a topic, I am typically not only interested in the topic in itself, but also in the context, the history, etc. If I read a citation, I quite often do not only want to learn about the contents of the work, but also about the circumstances when the work was written, in the publication history, in the author (education, life, philosophy), the publisher (history as a company, other publications), other related works on a subject by the same or other authors or publishers, etc. To me, this is interesting in itself (YMMV), but studying the background often enough has helped (me) to get a better insight into a topic, and it also helps to improve the contents and citations in Wikipedia. Wikidata is (or will be) a resource for semantic research beyond what Wikipedia provides, so IMO it is quite important that we establish a link to the corresponding entries there.

On a different matter, I do not in any way endorse the illegal distribution of works, but legally, it is not the act of pointing to, but the act of making available without appropriate license and the act of using of such material, that is illegal. By itself, pointing to is a neutral act, unless it would be deliberately misleading like advertising a link as a free download when the material is in fact not free. It is a (common) misconception to assume that anything that "opens" (without password) when clicking on it is free, and, by extension, that anything that is not free would be illegal to link to. By providing a link to Semantic Scholar or Wikidata, we do not assume responsibility for them.

Also, as Uzume has pointed out already, in the case of Wikidata, we at least have means to correct problems there. Whenever we do, we help to reduce the amount of incorrect information floating around in the net, and since (among others) Wikidata is used by information providers which in turn are used by our editors, this will indirectly also help to reduce the amount of crap that gets imported into Wikipedia (by refill, citoid, etc.) So, helping them, we indirectly help ourselves.

--Matthiaspaul (talk) 16:02, 17 July 2020 (UTC)

It's high time we have a Wikidata identifier (and possible one that's automatically appended when there's a matching doi/whatever). It's one database amongst many, but one we should highlight. If something is wrong on Wikidata, it's like anything on Wikipedia. Fix it. This is very different from {{cite Q}} which imports things from Wikidata, and which should never be done. Headbomb {t · c · p · b} 22:49, 8 July 2020 (UTC)

The problem with "so fix it" is that there is already plenty that needs fixing on Wikipedia without taking on responsibility for another project. Kanguole 10:59, 12 July 2020 (UTC)

I share most of the concerns regarding Wikidata (user interface, reliability, etc.), but nevertheless I think that it would be beneficial for both projects if semantically related entries are linked. For this, we do not have to take on responsibility for Wikidata. A link makes it easy to compare the data and possibly correct incorrect information, thereby reducing the amount of bogus info floating around in the net, which will indirectly also help us because auto-fill tools retrieving citation data from external resources will less likely import junk. Without a link, it is unlikely that errors will be detected until we would have an article about the title (which then has a link to Wikidata). Also, Wikidata's organization of data can aid further and deeper research in ways which would be out of scope for citations in Wikidata, but which nevertheless might be interesting for people researching a topic.

--Matthiaspaul (talk) 11:36, 13 July 2020 (UTC)

So we fill Wikidata with rubbbish and they fill WP with rubbish? I really do not see the point - it is just creating a massive circular mess. - Sitush (talk) 11:40, 13 July 2020 (UTC)

I wonder how you read that from my comment, because I stated more or less the exact opposite. Can we focus on solutions rather than polemics?

--Matthiaspaul (talk) 12:13, 13 July 2020 (UTC)

Another annual discussion. I'm afraid the position is uncharged. Wikidata still has reliability issues, possible copyvio traps, and circular reference/self-reference problems. CS1/2, which purports to be an aid to avoiding these, has a pretty good fix in place regarding Wikidata already. 65.88.88.69 (talk) 23:41, 8 July 2020 (UTC)

Of course Wikidata has issues. Everything does but we are already importing data from there for certain things (I doubt anyone is going to make sitelinks go away anytime soon). That said isn't it high time these things are addressed and made better? Commons also has its problems. I do not think it should just be written off because they exist. Copyvio problems are hard to answer. Sure there is the question if that should be in Wikidata but is it better for us to link to CiteSeerX, Semantic Scholar, or Google Scholar, etc. which has many of the same links? That seems pointless. I agree "fair use" is not context insensitive but public Internet links are, so these are copyvio and security concerns at the point of publication—as it has always been. Our linking to them or not isn't much different than using {{external media}} over Commons and local hosting. Are we policing all the CS1 |url= values and article external links too? At least one of the .edu links for Semantics of context-free languages (Q56672530) is also found at Attribute grammar#External links. If they are to be policed they should be equally so. Is it better to police them from here or a more consolidated location like Wikidata? My point is that Wikidata is WMF data and as such is our data. For that reason I believe it is foolish for WMF wikis to intentionally ignore such because it might need work. —Uzume (talk) 02:56, 9 July 2020 (UTC)

Just because we are already doing it for some things does not mean we should expand it to others. There are already numerous problems with the some things, such as short descriptions, and your argument is sort-of circular - "we already use it, so use it". I would oppose pretty much any further integration of Wikidata into this project until that project has got its act together, which looks likely to take years. If you think our syntax etc is mystifying, it is nothing compared to the experiment that is Wikidata and increased obscurity does not make things easier to work with. We are mostly writers, not data scientists. - Sitush (talk) 10:26, 9 July 2020 (UTC)

I support linking to Wikidata. So far, if we had a Wikidata entry but no Wikipedia article about a title or one of its authors, I have used |title-link=:wikidata:Q..., |author-link=:wikidata:Q... or |editor-link=:wikidata:Q... to establish a connection. If we have an article, there is no need to link to Wikidata directly, because the article (or redirect) will (or can) have a link to Wikidata instead. Bots running into iwls such as :<language>: or :wikidata: can follow the link and check if an article exists in the local Wikipedia meanwhile - if so, the link can be updated to point to the local article.

Obviously, this doesn't work for titles if the citation has a |url= as well. This is, where a |wikidata= parameter could be useful.

--Matthiaspaul (talk) 10:02, 9 July 2020 (UTC)

• To be fair, this isn't something we should be deciding with just a casual discussion. This should be a full RFC (which makes it painfully clear that we're not asking for a {{cite Q}}/{{cite DOI}} situation, but rather simply supporting links to WD). Headbomb {t · c · p · b} 14:15, 9 July 2020 (UTC)

• If someone does put together an RFC, please make the question something other than "should a |wikidata= parameter be added to CS1?" RFCs cause lots of trouble when they are poorly worded and incompletely thought out. If someone here is contemplating an RFC to decide this issue, please start a discussion first to clarify (for both potential supporters and potential opponents) what you are really asking for, functionally, as a change. You are more likely to get the result you want with a clear RFC statement and question. – Jonesey95 (talk) 14:54, 9 July 2020 (UTC)

Wow, I figured it would be good to ask here first but I really did not think asking for the addition of a single identifier for a WMF sister project would required extended discussion (not that I am against such). I wonder if it would take as much for other potential identifiers like Semantic Scholar corpus ID (P8299) or ResearchGate publication ID (P5875) (I am not suggestion they be auto-populated by Wikidata just CS1 parameters with arguments with the same well defined identifier meanings; any potential auto-population can be saved for the likes of {{cite Q}}, etc.). Those can and do store/cache actual copies of scholarly articles much like CiteSeerX does (although I personally feel Semantic Scholar (Q22908627) is the best of the three). —Uzume (talk) 17:00, 9 July 2020 (UTC)

It is likely that you have a clear vision of what you are asking for, but some of us here do not. Please give an example of how this proposed new parameter would work. Show us what an example cite template would look like in code and rendered. You don't have to actually code the template, just show us a mockup of the rendered wikicode. – Jonesey95 (talk) 17:17, 9 July 2020 (UTC)

It'd be something like

• Smith, J. (2015). "Title of Things". Journal of Stuff. 34 (1): 23–45. doi:10.4321/3210. PMID 012345. Wikidata:Q0123465.

Headbomb {t · c · p · b} 17:38, 9 July 2020 (UTC)

Thanks, that helps. So the QID would point to a Wikidata entry for the item described in |title= for {{cite journal}}? Is that the only template in which |wikidata= would be supported? Are there a lot of useful Wikidata entries for journal articles? Can you provide a real-world example with a real article? I poked around at Wikidata for a bit, but none of my normal Wikipedia tricks like "What links here" work as I expect over there, so I was unable to find a transclusion count, or whatever it is called at Wikidata. – Jonesey95 (talk) 18:19, 9 July 2020 (UTC)

That would be general parameter for for all CS1/CS2 templates, no different than |bibcode= or |pmid=. Headbomb {t · c · p · b} 19:04, 9 July 2020 (UTC)

Since |wikidata= is a bit ambiguous and could mean a lot of things (some of which some people seem to be afraid of), perhaps it would make sense to name the parameter |qid= to make sure that we are linking only to item node IDs, not property IDs - this would also be more in line with the other parameters for identifiers, where we used the abbreviated name. Also, to make it impossible to link to other stuff, the "Q" prefix could be made part of the predefined part of the link, so it would look like:

• Smith, J. (2015). "Title of Things". Journal of Stuff. 34 (1): 23–45. doi:10.4321/3210. PMID 012345. QID Q0123465.

or even just

• Smith, J. (2015). "Title of Things". Journal of Stuff. 34 (1): 23–45. doi:10.4321/3210. PMID 012345. QID 0123465.

If this still isn't short enough to be unobtrusive, we could even think about reducing this to something like:

• Smith, J. (2015). "Title of Things". Journal of Stuff. 34 (1): 23–45. doi:10.4321/3210. PMID 012345. QID @.

After all, the QID number will be used only to establish a connection between the reference and the Wikidata entry, not to cite the work externally, so it is not really necessary to display the ID number in the citation for as long as there is some element a user can click on to jump to the Wikidata entry.

--Matthiaspaul (talk) 10:53, 12 July 2020 (UTC)

@Matthiaspaul: Being so short, I think "QID" is also too ambiguous and it has already been used by others (e.g., What is a QID identifier?). Why not |wdqid= which is inline with |s2cid= which CS1 already has (as a fairly recent addition)? It should perhaps be noted at this point that the QID is not really a Wikidata thing so much as a Wikibase thing. So far there are no other significant Wikibase databases besides Wikidata but nothing stops anyone from using Wikibase for something else in the future (e.g., Commons already has another but so far they are only using specialized MediaInfo entities and otherwise leveraging Wikidata item and property entities). Talk:WDQ (identifier) is also some related discussion on this topic. —Uzume (talk) 03:46, 17 July 2020 (UTC)

|wdqid= would be another possibility. IMO, it is better than |wikidata=, but not as good as |qid= - unless there would be other work or author identifiers named QID in use somewhere:

The link you provided lists one other identifier named QID, but it is not used in the context of citations. So far, I couldn't find any information on Commons' regarding how they call their IDs, but unless they would be called QIDs as well, I don't see a problem there as well - we would not link to both Wikidata and Commons, because the link to Commons could be found at Wikidata.

Although this is not a hard requirement, ideally, the name of the parameter would match the abbreviated (official) name of the identifier for consistency, that's why I (still) prefer |qid=, but would support |wdqid= as well, of course.

--Matthiaspaul (talk) 10:19, 18 July 2020 (UTC)

Can someone explain to me what useful information a wikidata link on a reference would provide to a reader of our articles? Because I'm not seeing it. It just looks like clutter to me. If we were storing reference metadata on wikidata and using a template parameter to tell the template code to expand the metadata from there, that would be one thing, but that's not what's being proposed. I don't see the value in making wikidata ids visible to readers. —David Eppstein (talk) 19:05, 9 July 2020 (UTC)

Things like author affiliation, ORCIDs, possibly other (more obscure) identifiers like holdings/links to specific libraries (e.g. BnF) etc... Crossed-linked to information about the journal (e.g. ISSN, OCLC) and so on. It's not the most useful of things, but I'd consider it way more useful than an ISSN link personally. Headbomb {t · c · p · b} 20:00, 9 July 2020 (UTC)

The point of a citation is to enable the reader to find the source of the information should they want to. I'm with David Eppstein here; a wikidata link would provide no useful information to the reader about the citation. (An ISSN link isn't usually useful, I agree, but can very occasionally clarify which journal is meant when there are journals with very similar names or names which have changed over time.) Peter coxhead (talk) 20:16, 9 July 2020 (UTC)

(edit conflict) If an editor thinks |issn=, |oclc=, or |isbn= are needed to identify/disambiguate/help locate a particular source, they're already going to add them to the citation. I'm not why it's useful to send readers to another site in order to get yet another, more obscure or less-useful identifier. IDs like ISSN/OCLC/ISBN are useful outside of the Wikimedia Foundation; if someone has a printed out copy of a Wikipedia article, they can use those IDs to fill out, an interlibrary loan request form, to search a library catalog, etc. I realize Wikipedia is an online encyclopedia and hence most viewers will be able to click on links (such as the present system of links from OCLCs and ISSNs to Worldcat and from ISBNs to Special:BookSources, which seems more than sufficient to help readers locate sources), but linking Wikidata entries just seems like extra clutter for not much added benefit. Citations are to help readers and editors find the source; being able to find fields like author affiliations and the like seem superfluous to me. Umimmak (talk) 20:31, 9 July 2020 (UTC)

I usually omit or remove ISSN identifiers for the same reason, unless it's a particularly obscure periodical whose identity might not be clear: they don't help readers locate the specific publication being cited. An ISBN at least goes to the specific book being cited, and can lead to online resources like Google Books or WorldCat, so those are more useful. —David Eppstein (talk) 20:36, 9 July 2020 (UTC)

...but ISSNs can also lead to WorldCat, which quickly lists libraries where issues of the periodical are available? I don't understand that part of your comparison. Glades12 (talk) 07:00, 10 July 2020 (UTC)

It should be noted the Wikidata Q identifier (Q43649390) can be used to link to other things besides the entity page, e.g., d:Special:EntityPage/Q56672530 vs. Q56672530 @ Scholia (Q45340488) vs. Q56672530 @ Reasonator (Q20155952), etc. In my opinion some are considerably less "noisy" than others and even better ones could be built for citation purposes (if they do not already exist). —Uzume (talk) 05:28, 17 July 2020 (UTC)

Another approach for Wikidata links

Thinking about how to integrate support for linking to Wikidata in a way acceptable to most users, I would like to discuss an alternative approach, making such links as unobtrusive as possible (and even conditional on user opt-in) by just providing a small clickable Wikidata symbol instead of listing them as "Wikidata:Q123456" or "QID Q123456" among the other identifiers (however, that would be possible as well). Commons does this in a similar way already, see f.e. in the "Summary" section here: c:File:Tarzan_and_the_Golden_Lion_-_McClurg1923.pdf

Rationale:

• Wikidata is our sister project; this might warrant to special case them somewhat compared to other identifiers.
• Wikidata QIDs are public and mostly static, but they are not designed to be semi-permanent. In contrast to other identifiers, they are not used in the outer world to refer to works in citations. Therefore, it is not actually necessary to present those Q numbers in a citation for as long as we can establish the link.
• As soon as we have an article about a title (or an author) we do no longer need to link to Wikidata directly, as the corresponding entry can be reached through the "edit" link for inter-wiki links in an article's side bar. Therefore, direct links to Wikidata are often (although not always) a temporary measure in the process to set up infrastructure.
• At least I find it desirable to establish links to Wikidata for authors even more than for works, because author names are often ambiguous and therefore linking entries can help to determine a particular author and avoid possible mix-up of the names in the future (can also be used to help automate the process of establishing iwls).
• Introducing a dedicated parameter like |qid= or |wdqid= can help to simplify linking to Wikidata entries for works, but not for authors. In general, using |-link= parameters with :d: or :wikidata: prefixes is a more flexible approach, with the sole exception of a |url= external link to a work being provided as well. In this case, the two links should both be given (instead of disallowing this combination as we do when the |title-link= would not point to Wikidata), with the Wikidata link presented by a Wikidata icon immediately following the external link, only. (If it would be desirable to support |-link= links to Wikipedia articles in addition to links to Wikidata, a dedicated parameter like |qid= or |wdqid= would still be needed.)

Thereby we could integrate Wikidata links in a very "natural" way, and even indicate to users that the link goes to Wikidata before they have to click the link (per the principle of least surprise and with all implied disclaimers regarding user interface, reliability, etc.). The approach would avoid to actually display the Q number or other extra clutter like "Wikidata:Q123456" or "QID Q123456" in the citation, and links to Wikidata entries for works and authors would work in an identical way.

Example 1:

{{cite book |last=Burroughs |first=Edgar Rice |date=1922 |title=Tarzan and the Golden Lion |title-link=:d:Q9081967 |publisher=A. C. McClurg |page=3}}

Burroughs, Edgar Rice (1922). Tarzan and the Golden Lion . A. C. McClurg. p. 3.

Alternative implementation of example 1: If the title would not be included in the link, the icon labels to Wikidata could even be made conditional on user opt-in through CSS, so that they become available only to more advanced users:

Burroughs, Edgar Rice (1922). Tarzan and the Golden Lion . A. C. McClurg. p. 3.

Example 2: Special case if both |url= and |title-link= to Wikidata would be given:

{{cite book |last=Burroughs |first=Edgar Rice |date=1922 |title=Tarzan and the Golden Lion |title-link=:d:Q9081967 |publisher=A. C. McClurg |page=3 |url=https://books.google.com/books?id=Wrk8AAAAYAAJ}}

Burroughs, Edgar Rice (1922). Tarzan and the Golden Lion. A. C. McClurg. p. 3.

--Matthiaspaul (talk) 12:49, 18 July 2020 (UTC)

What percentage of readers would recognize the Wikidata logo and understand that it is a clickable link? How would this usage meet MOS:ICON? Nikkimaria (talk) 14:16, 18 July 2020 (UTC)

At least the people at Commons seem to assume that the Wikidata icon is already recognized widely enough for using it in their database interface templates (see the Commons link above).

In the first variation of the first example above, the citation title is a clickable blue link as well, with just the Wikidata icon added afterwards. In that case there is almost no difference compared to other icons (like those "external link", "access lock" or "PDF" icons defined here ([5]), in fact, this would be better than the current state of affairs where |-link= links show as internal links without any special link decoration, so the users won't know that they will jump to Wikidata before clicking the link.

As shown in the second example, only in the case that we would have to provide a |url= external link in addition to a |title-link= link to Wikidata, the icon would become a separate element by itself without some label text (except for the alt= text), but still next to the citation's title. (Same for the alternative proposal to support a |-link= link to Wikipedia and a |qid= link to Wikidata in parallel - personally, I think, we can treat links to internal articles and links to Wikidata as mutually exclusive, but others might not agree on this.) An alternative would be to list the Wikidata link among the other identifiers as in the other examples further above. (I support both approaches, but the community would have to agree on only one of them, of course.)

In this special case of providing two links, some users may, for a while, miss the fact, that clicking on the blue title would jump to the external URL or Wikipedia's article and clicking on the icon would transport them to the corresponding Wikidata entry instead, but that is something than can be explained in the documentation, and the users would probably discover this by themselves as well. While it would be prohibitive to use a graphical-only element for some essential user interface, I do consider a graphical-only approach acceptable in this particular case because this scenario occurs only when two links are provided at the same time, and because the Wikidata link is a non-essential element. The icon is embedded in the normal text flow of a citation, instead of being some free-floating picture. Therefore, it can be selected even with text-only browsers and screen readers. They would either present the alt= description "Wikidata link to Q9081967" as link label or render the link similar to something like [https://www.wikidata.org/wiki/Q9081967] or [I]. So, the feature is usable even by visually impaired people.

Finally, if we present these icons only after opt-in (even better would be to display them by default and mute them after opt-out, if that's technically possible), this would address any concerns regarding it being a non-recognized or "unwanted" icon.

--Matthiaspaul (talk) 18:04, 18 July 2020 (UTC)

Following up on this, I meanwhile found {{Cite wikisource}}, a citation template I was unaware of. It is interesting in this context because it does something very similar for links to Wikisource already (although in a dedicated template, and not for dual title links) by appending the Wikisource icon as link decoration to Wikisource links:

{{cite wikisource |editor-first=Hugh |editor-last=Chisholm |chapter=Aard-vark |wslink=1911 Encyclopædia Britannica |plaintitle=[[Encyclopædia Britannica Eleventh Edition|Encyclopædia Britannica]] |edition=11 |date=1911 |publisher=Cambridge University Press}}

Chisholm, Hugh, ed. (1911). "Aard-vark" . Encyclopædia Britannica (11th ed.). Cambridge University Press – via Wikisource.

Since links to Wikisource are on a middle-ground between internal and external links, I think, per the principle of least surprise, we should incorporate this idea into the other citation templates as well by appending the icon whenever a |-link= parameter points to Wikisource, that is, when its value is prefixed by :s: or :wikisource:.

And, by extension, the same should happen for links to Wikidata, when a :d: or :wikidata: prefix is used in a |-link= parameter.

--Matthiaspaul (talk) 20:45, 19 July 2020 (UTC)

See also: #Proposal_to_support_link_decoration_not_only_for_links_to_Wikisource_but_also_for_other_inter-wiki_links

--Matthiaspaul (talk) 13:29, 9 August 2020 (UTC)

Wikidata is our sister project, therefore it reflects worse on us than other sites might when they host pirate links and we boost their signal in doing so. Does Wikidata even have any guidelines or policies about which external links are or are not permissible? When I follow the wikidata link from WP:EL to its equivalent page on Wikidata, I get a page https://www.wikidata.org/wiki/Q4657623 with no guidelines and no discussion. Maybe that is only the wikidata item hosting cross-links to different sites' external link guidelines and not the guidelines themselves, but it has no cross-link to Wikidata itself. —David Eppstein (talk) 18:38, 18 July 2020 (UTC)

You already linked to the discussion of property "full work available at URL" (P953) above. I meanwhile found what appears to be the original proposal to add this property: https://www.wikidata.org/wiki/Wikidata:Property_proposal/Archive/15#P953

From this it can be deduced that the links are meant to point to legal free or non-free resources (including links pointing to resources behind a paywall). Pointing to openly accessible illegal copies of a work are not an intended use, so such links could be safely deleted when spotted.

My point is that this is an issue that can't be solved by ignoring it, in particular not ignored by us as Wikipedia is already interconnected with Wikidata in many ways, and (we may like it or not) the WMF is enforcing an even tighter integration in the future. So, if Wikidata looks bad, this already reflects bad on us. The only solution, as I see it, is to fix the issues at Wikidata as soon as possible. Providing links to Wikidata, as proposed in this thread, can IMO only help to put the focus on the corresponding Wikidata entries which are relevant for us and clean them up if necessary. As most citations at Wikipedia are probably better researched than Wikidata entries, Wikipedia's citations could become part of the necessary web of trust for a more reliably future Wikidata. Without a link, problems may continue to exist there for many years to come, which will fall back bad on us.

However, in all those years I am linking to Wikidata through |-link= parameters, I cannot remember a single case where the Wikidata entry contained links to illegal copies of a work. In most cases, the Wikidata entries didn't use P953 at all. If that would have been the case, I would probably have deleted the questionable links there - or would not have provided a link to the Wikidata entry in the first place.

Just because we would have integrated support for Wikidata links we are not obligated to provide such links if there is anything questionable about an entry (and cannot be fixed) there. This is not different from links provided via |url= - if we find one which is problematic, we would remove it from a citation.

So, arguing against links to Wikidata just because some entries over there are problematic despite the many other entries without issues seems like throwing the baby out with the bathwater to me. To make sure that such links would not be abused, we could add a reminder to the documentation that links to problematic entries should be removed. --Matthiaspaul (talk) 23:01, 18 July 2020 (UTC)

@Matthiaspaul: I like the idea and the logo is well associated with Wikidata even locally (see this search), however, I wonder how this might fly in the face of current policy with regard to icons being the main link target label in main article space, e.g., see {{Wikidata icon}} and and the results of its TFD: Wikipedia:Templates for discussion/Log/2018 January 16#Template:Wikidata icon. On the other hand, we also have things like {{EditAtWikidata}} (albeit with different icon) which is used extensively in main article space as the main link target label for Wikidata links. Perhaps the "edit at" pencil (which is specially locally cached, see File:OOjs UI icon edit-ltr-progressive.svg and Wikipedia:Cascade-protected items#Cascade protected images) would be better specialized with the Wikidata logo behind it (e.g., File:Wikidata CheckUser.svg, the Wikidata logo with a magnifying/search glass over it and of course File:Wikidata-edit.svg). The icon used could also further be specialized by having a default icon sizing such as File:Notification-icon-Wikidata-logo.svg. —Uzume (talk) 20:17, 19 July 2020 (UTC)

Thank you very much for the links, I wasn't aware of these templates and prior discussions.

I would have no problems using a different icon, if that would improve acceptance (I particularly like the loupe). The pencil icon might be problematic, though, because it somehow implies that we would pull Wikidata contents into the local citation, which we expressively do not want to do in this proposal.

--Matthiaspaul (talk) 21:22, 19 July 2020 (UTC)

Date ranges with seasons

I've suddenly seen several cite errors for citations with dates that use season ranges, for example in Franz Kafka:

• Panichas, George A. (Spring–Fall 2004). "Kafka's Afflicted Vision: A Literary-Theological Critique" (PDF). Humanitas. 17 (1–2). Bowie, Maryland. {{cite journal}}: Invalid |ref=harv (help)

These were not edited recently, so why would any recent change make this a new error? — Preceding unsigned comment added by Kennethaw88 (talk • contribs)

This looks like a bug possibly introduced by the work done for the addition of quarters that was made. Trappist the monk looks to be working on it. --Izno (talk) 00:20, 12 July 2020 (UTC)

I'm about to turn into a pumpkin so I'll get to it tomorrow morning. I know the cause, the issue is how to best fix it.

—Trappist the monk (talk) 00:24, 12 July 2020 (UTC)

Fixed, I think. Live module updated.

—Trappist the monk (talk) 12:00, 12 July 2020 (UTC)

Template-specific help in preview

Hi, I guess most of us do not remember all parameters supported by the various citation templates, in particular those which are not generic, but specific to certain templates.

In order to make it easier (quicker) to look up the template specific help page, I propose to let the template display an unobtrusive link to its help page (only) in article preview, either in front of the citation or after it.

This could look like

{{Cite journal}} would have a link to Help:Cite journal, {{Cite conference}} to Help:Cite conference, {{Cite book}} to Help:Cite book etc.

If even this small [?] would be found to be too obtrusive, it could be put into some CSS stuff so that it would show only when an editor has opted in to maintenance messages.

--Matthiaspaul (talk) 11:01, 13 July 2020 (UTC)

• I'm not sure I'd support that specific implementation, but the idea has merit and I'd support some variant of it. Possibly

As a preview/opt-in message. Headbomb {t · c · p · b} 12:42, 13 July 2020 (UTC)

Whatever we can agree upon. I deliberately tried to make it as short and unobtrusive as possible to not change the general appearance in preview, but if people would prefer longer messages as in your example, I would not be against it, either. --Matthiaspaul (talk) 12:54, 13 July 2020 (UTC)

Perhaps something like this (image instead of text) would be aesthetically more pleasing?

--Matthiaspaul (talk) 15:54, 13 July 2020 (UTC)

Really should be at the end I feel, but YMMW. The real question is do we want this as default, or as an opt-in preview, or handled by a separate script? Headbomb {t · c · p · b} 18:45, 13 July 2020 (UTC)

I definitely don’t think it should be the default. It’s confusing for editors to see things in the preview that don’t appear in the finished page, and question marks or comments like “need help?” suggest something is questionable or that editor needs help, implies something is wrong even when it’s formatted correctly. Umimmak (talk) 18:58, 13 July 2020 (UTC)

This seems like something that could be created as a javascript that would give the little icon or a link for all templates, not just CS1 templates. I see people having trouble with editing a wide variety of templates, putting in incorrect parameters and much more. – Jonesey95 (talk) 19:02, 13 July 2020 (UTC)

Nothing against a script-based solution, but unfortunately I personally couldn't take much advantage of it then as I usually have JavaScript disabled for security reasons...

--Matthiaspaul (talk) 20:55, 13 July 2020 (UTC)

I don't think this would cause much confusion, as it is self-explanatory. The curious user would click it once or twice to see what it is, and then take it for granted almost as if it would be a style element of the skin. I don't think the user needs a lengthly description in each citation, because once read its purpose is obvious. Some users might even miss the feature for some while, but that's the same with other interface elements - they will be stumbled upon and then used (if intuitive). That's why my preferred implementation would use a link as small as possible - ideally, I would even "reuse" some existing style element (like the "^"), but then it would have to be implemented as part of Mediawiki's <ref> token rather than inside our local citation templates (however, this won't work because only the citation template itself knows the location of its help page).

If a question mark would draw the wrong association, we could use something like a "book" or "file" icon to indicate "documentation" or even something more abstract like a (diagonally) upwards-pointing arrow (similar to the external-link icon, but pointing in one direction only) to indicate "look up".

--Matthiaspaul (talk) 20:55, 13 July 2020 (UTC)

The reason why I prefer the link in front of the citation (at least when it remains as short as in my example) is because it then "blends in" visually with the other "strange" symbols (like the "1./2.", "^" typically found in front of citations), whereas at the end of a citation there is often other text following (inside the <ref> block), and it might look "out of place" there or actually cause confusion (when short).

If we'd use a long descriptive text label, we would definitely need to frame it for opt-in. If it remains just an icon (or similarly short) this would not be needed (thereby we'd include the large group of "normal" users who don't change their default configuration).

--Matthiaspaul (talk) 20:55, 13 July 2020 (UTC)

Following up on icon prefixes, I just ran into {{cite wikisource}}, a cite template I wasn't aware of. This actually uses a similar format already:

 Chisholm, Hugh, ed. (1911). "Aard-vark". Encyclopædia Britannica (11th ed.). Cambridge University Press – via Wikisource.

So, with a question mark icon added in preview mode this would look similar to:

Not too bad, I would think...

--Matthiaspaul (talk) 16:20, 19 July 2020 (UTC)

Nice coincidence, when one clicks on "View history" on a page, a circled black question mark icon appears in the upper right corner leading to Help:Page history. So, there is even some precedent for this... ;-)

--Matthiaspaul (talk) 20:58, 19 July 2020 (UTC)

cs1|2 renders differently in preview mode only when there are archive-url errors. The initial implementation of that queried the {{REVISIONID}} magic word for every citation whether there were errors or not. That implementation drew the attention of MediaWiki because it took longer to save pages. This because preview-pages are different from saved pages so MediaWiki can't reuse the preview and must parse the wikitext again before it can be saved. See Help talk:Citation Style 1/Archive 20 § archive url checks and preview mode and the related discussion at Wikipedia:Village pump (technical)/Archive 147 § Preview-only template warnings using REVISIONID magic word.

—Trappist the monk (talk) 19:17, 13 July 2020 (UTC)

Obviously, I would have used {{REVISIONID}} as well to detect preview mode... Let's loop in User:Aaron Schulz to see if there is meanwhile another way to detect preview - after all, four years have passed...

Is there a CSS class for "preview messages", that is, something that is visible only in preview without being conditional on {{REVISIONID}}?

--Matthiaspaul (talk) 02:22, 14 July 2020 (UTC)

A bit too much insider jargon for me to decipher if the issue has been worked around by User:Aaron Schulz in 2019, so I'll drop these finds here for evaluation and comment by those who are familiar with the implementation:

• https://gerrit.wikimedia.org/r/c/mediawiki/core/+/294774/ 2019-04-02 "Disable expensive {{REVISIONID}} magic word in miser mode"
• https://phabricator.wikimedia.org/T137900 "Deal with poor edit stash hit rate due to Lua modules using {{REVISIONID}}"
• https://gerrit.wikimedia.org/r/c/mediawiki/core/+/503685/ 2019-04-15 "parser: use "-" for revision ID for non-preview edit filter parse during save"
• https://phabricator.wikimedia.org/T235957 2019-10-19 "Change {{REVISIONID}} from number to "-" in wgMiserMode"
• https://gerrit.wikimedia.org/r/c/mediawiki/core/+/570985/ 2020-02-18 "parser: apply $wgMiserMode restriction to REVISIONID for NS_USER/NS_PROJECT"

--Matthiaspaul (talk) 17:26, 15 July 2020 (UTC)

On 2019-04-04, User:Krinkle wrote ([6]):

"The magic word {{REVISIONID}} is being deprecated for performance reasons. In the future, it will only return "" (empty string) when previewing edits, or "-" (dash) when reading pages. The release of next week, will only change this behaviour for articles in the content namespaces. It will not change for interface messages and other namespaces (such as talk pages, and user pages)."

So, it basically has reduced to a flag now.

Anyway, I would prefer to see such help links in preview rather than saving a few seconds on save. Comments?

--Matthiaspaul (talk) 15:49, 18 July 2020 (UTC)

Having reread the other threads, I have come to the conclusion that we can assume the issue as being worked around at least to an extent that using {{REVISIONID}} does no longer cause a significant performance hit that could not be remedied on server side.

On 2016-06-22, User:Ori Livneh/User:ATDT wrote (Help talk:Citation Style 1/Archive 20#archive url checks and preview mode):

We try to do all the work except actually save the edit to the database. If the user does not end up saving the edit or if the user goes back and makes additional changes, we just throw away whatever we computed. But if the user saves the edit, then often times a lot of the work is already done and all we have to do is commit it to the database. Whenever the REVISIONID magic word is used, this whole mechanism is basically subverted, because we can't reliable know in advance what the revision ID is going to be before we save it to the database.

However, with {{REVISIONID}} reduced to something like a {{PREVIEWMODE}} flag, the pseudo-revision ID can be predicted (even during preview) to be "-" when later saving a page, thus the precompiled page can be reused with no or only minor fixups. What still might differ is the resulting HTML (with or without the help icons), but nobody would keep them from silently generating the HTML for the "-" case during preview already (in addition to the preview itself), so, while this may still require two passes, there is no need to defer the final pass until the user hits "Save", thus no performance penalty on the user side.

In the worst case, the feature could be made conditional on citation errors occuring. In this case, it would not show for all citations in preview but at least for those where errors were detected and thus help is particulary useful.

--Matthiaspaul (talk) 18:06, 19 July 2020 (UTC)

Apparently, none of the three pinged server admins/developers can be bothered to reply. Since the problem apparently no longer exists, let's go for it. --Matthiaspaul (talk) 12:35, 8 August 2020 (UTC)

Better URL error detection

• Chrissi-Yianna Politou; Konstantinos Kapiris; Porzia Maierano; Francesca Capezzuto; John Dokos (2004). [https:www.vliz.be/imisdocs/publications/68724.pdf "Deep-sea Mediterranean biology: the case of Aristaeomorpha foliacea (Risso, 1827) (Crustacea:Decapoda:Aristeidae)"] (PDF). Scientia Marina. 68 (Supplement 3): 129–139. doi:10.3989/scimar.2004.68s3129. {{cite journal}}: Check |url= value (help)

This should throw an error. Headbomb {t · c · p · b} 21:06, 13 July 2020 (UTC)

Fixed in sandbox? There are uris that do not use the authority indicator (//); news: is one such that cs1|2 supports. I have used that one as a test: if scheme is not news: then authority indicator is required.

{{cite journal/new | author1 = Chrissi-Yianna Politou | author2 = Konstantinos Kapiris | author3 = Porzia Maierano | author4 = Francesca Capezzuto | author5 = John Dokos | year = 2004 | title = Deep-sea Mediterranean biology: the case of ''Aristaeomorpha foliacea'' (Risso, 1827) (Crustacea:Decapoda:Aristeidae) | url = https:www.vliz.be/imisdocs/publications/68724.pdf | journal = Scientia Marina | volume = 68 | issue = Supplement 3 | pages = 129–139| doi = 10.3989/scimar.2004.68s3129 }}

Chrissi-Yianna Politou; Konstantinos Kapiris; Porzia Maierano; Francesca Capezzuto; John Dokos (2004). [https:www.vliz.be/imisdocs/publications/68724.pdf "Deep-sea Mediterranean biology: the case of Aristaeomorpha foliacea (Risso, 1827) (Crustacea:Decapoda:Aristeidae)"] (PDF). Scientia Marina. 68 (Supplement 3): 129–139. doi:10.3989/scimar.2004.68s3129. {{cite journal}}: Check |url= value (help)

—Trappist the monk (talk) 22:41, 13 July 2020 (UTC)

Bogus long volume

The following reference generates a "CS1: long volume value" error, but the volume is correct. We need to accept volumes of reasonable length with a hyphen or en dash.

{{Cite journal |last=Croitor| first=Roman |last2=Stefaniak |first2=Krzysztof |last3=Pawłowska |first3=Kamilla |last4=Ridush |first4=Bogdan |last5=Wojtal |first5=Piotr |last6=Stach |first6=Małgorzata |date=April 2014 |title=Giant deer ''Megaloceros giganteus'' Blumenbach, 1799 (Cervidae, Mammalia) from Palaeolithic of Eastern Europe |url=https://linkinghub.elsevier.com/retrieve/pii/S1040618213008689 |journal=Quaternary International |language=en |volume=326–327 |pages=91–104 |doi=10.1016/j.quaint.2013.10.068}} : Croitor, Roman; Stefaniak, Krzysztof; Pawłowska, Kamilla; Ridush, Bogdan; Wojtal, Piotr; Stach, Małgorzata (April 2014). "Giant deer Megaloceros giganteus Blumenbach, 1799 (Cervidae, Mammalia) from Palaeolithic of Eastern Europe". Quaternary International. 326–327: 91–104. doi:10.1016/j.quaint.2013.10.068.

Please let me know when this is fixed. —Anomalocaris (talk) 00:14, 15 July 2020 (UTC)

Not an error. What makes you think that it is?

—Trappist the monk (talk) 00:26, 15 July 2020 (UTC)

Long volume is not something to be corrected at this time; that's why it is listed neither as an error nor as a maintenance category but instead as a property. We have not had a conversation since instituting the check for these pages whether we want to do anything about them. --Izno (talk) 01:39, 15 July 2020 (UTC)

If this was an error, the text about it would have been red on most devices, like in the following example: . Vanity Press [go to the library go to the library] {{cite book}}: |first= missing |last= (help); Check |url= value (help); Missing or empty |title= (help). Glades12 (talk) 06:44, 15 July 2020 (UTC)

Trappist the monk: Do you mean:

• "I don't see the problem, please explain it."
• "It's not an error, it's supposed to treat |volume=326–327 as a long volume."
• "For an article to be in Category:CS1: long volume value is not a sign of any actual error."
• Something else ...

My point is that because of the citation given here, and for no other reason, Irish elk is included in Category:CS1: long volume value, even though this volume is correct, and this category is an error category. —Anomalocaris (talk) 10:53, 15 July 2020 (UTC)

I mean that what you have written (generates a "CS1: long volume value" error and this category is an error category) is not correct. Category:CS1: long volume value is not an error category but is a properties category. What is it, or what have you read, that makes you think that the category is an error category? If it is something that you read, we can fix that when you tell us where you read it.

—Trappist the monk (talk) 11:25, 15 July 2020 (UTC)

What we (why did you only ping TPM?) mean is that this property may constitute an error, but is not one in all cases. Thus, it is separated from the error categories because the latter always contain real errors that should be corrected in any case. This is intentional, and you are correct in that 326–327 is a real volume. Glades12 (talk) 11:56, 15 July 2020 (UTC)

To reiterate, Category:CS1: long volume value is not an error category, and neither you nor the creators of it made any mistake here. Glades12 (talk) 12:01, 15 July 2020 (UTC)

Anomalocaris, is there something about the text at Category:CS1: long volume value that is confusing? I tried to be as clear as possible when I wrote it, but if it can be improved, please improve it or let us know how it was confusing. – Jonesey95 (talk) 13:25, 15 July 2020 (UTC)

Template to warn that other template is not safe for use in citations

Where's the template that we use in the /doc of various utility templates (I'm having trouble remembering a specific one) that warns people not to use that tagged template inside a citation template, because it pollutes the COinS metadata?  — SMcCandlish ☏ ¢ 😼  00:27, 15 July 2020 (UTC)

@SMcCandlish: {{COinS safe|n}}, as seen on {{Interlanguage link}}. – Finnusertop (talk ⋅ contribs) 00:37, 15 July 2020 (UTC)

Thankee. That's just what I was looking for.  — SMcCandlish ☏ ¢ 😼  03:38, 15 July 2020 (UTC)

Emoji HTML help

I know there's been a previous help issue located here, so my quick question/help is how to get the emoji from this tweet working since I'm also getting the zero-width joiner error. I've tried looking up the different HTMLs for emojis, but was unable to fix it. The tweet is currently cited on this article. Thanks in advance! Magitroopa (talk) 17:27, 15 July 2020 (UTC)

I've fixed those this way:

1. copy the emoji
2. go to this tool
3. at the right, paste the emoji into the 'text area' box → 👨‍🍳
4. click the A># button below the text area → 1F468 200D 1F373
5. delete 200D (zero width joiner character) → 1F468 1F373
6. click the #>A button below the text area → 👨🍳
7. insert ‍ html entity between the two emoji → 👨&zwj;🍳
8. replace the emoji in wikitext with the two-emoji-and-html-entity string:

Crude example; the original emoji:

{{cite web |title=👨‍🍳 |url=//example.com}}

"👨‍🍳".

the modified emoji:

{{cite web |title=👨‍🍳 |url=//example.com}}

"👨‍🍳".

—Trappist the monk (talk) 17:54, 15 July 2020 (UTC)

Dead link in a Featured Articles

I have tried to implement this diff] but it is being reverted for reasons: the article is a featured article, FA are not allowed to have dead links in them (!), so the archive.org is placed in the |url= field so that it is not apparent the link is dead (!!) thus ensuring the FA won't get delisted (!!!). I have tried to explain to User:Neutralhomer that this is nothing to be concerned about, but have not been successful. Turning it over to the community should anyone be interested in helping Neutralhomer and correcting this citation. -- GreenC 01:55, 16 July 2020 (UTC)

Really? On the day of its promotion, that article had four {{cite web}} templates that use |archiveurl= and |archivedate= (78, 79, 80, 85). It appears that those same four are in the article as I write this (84, 85, 86, 91). Neither of the terms archive nor dead appear in Wikipedia:Featured article criteria.

If the objection is to the inclusion of |url-status=dead, that can be omitted (when any value is assigned to |archiveurl=, cs1|2 assumes that |url= is dead).

Certainly the use of |archive-url= and |archive-date= is normal day-to-day business with cs1|2 templates.

—Trappist the monk (talk) 02:40, 16 July 2020 (UTC)

Neutralhomer is in the wrong here. A proper citation much include the original link, marked as dead, and give the archived link (with the archive date). Headbomb {t · c · p · b} 03:21, 16 July 2020 (UTC)

Publication date

Is there a different validation for |publication-date= as it is reporting an error which does not happen if it is |date=?

Rik Farrow (2018). Rik Farrow (ed.). "Musings" (PDF). ;login. 43 (4). USENIX (published Winter 2018): 4. ISSN 1044-6397. {{cite journal}}: Check date values in: |publication-date= (help)</ref>

Keith D (talk) 21:17, 17 July 2020 (UTC)

Yes. Seasons, named dates (Christmas, Easter), and quarter dates are only supported by |date=.

—Trappist the monk (talk) 21:33, 17 July 2020 (UTC)

Makes sense to support it in |publication-date= as well, I guess. --Matthiaspaul (talk) 11:20, 18 July 2020 (UTC)

Thanks. May be documentation could include a note to explain. Keith D (talk) 12:02, 18 July 2020 (UTC)

I would just make |publication-date= an alias of |date= and deprecate the long alias or remove it from the docs; we do not need to "advertise" every alias that has ever existed.  — SMcCandlish ☏ ¢ 😼  13:04, 4 August 2020 (UTC)

Request: Please make it so |lang=zh-tw displays "Taiwanese Mandarin" and not "Chinese"

Before

• 回首頁 [Home page]. 臺灣閩南語常用詞辭典 [Dictionary of Frequently-Used Taiwan Minnan] (in Chinese (Taiwan) and Taiwanese Hokkien). Retrieved 2020-06-28.

After

• Template:Cite web/sandboxHokkien

Proposed diff

[7] (without line 76 for now, will be separate request if this one approved)

Rationale

MediaWiki's name for zh-tw makes no sense. It outputs "Chinese" for zh-tw, zh-cn, etc. ISO really dropped the ball for zh. Languages as diverse as Cantonese, Beijing dialect, and Wu Chinese can be, and often are, all called "Chinese" even though they are not at all mutually intelligible in speech, especially traditional forms. (Nota bene: Varieties of Chinese.)

There's a huge political dimension to all this that there simply is not when it comes to calling en-US and en-UK both English. The Chinese Communist Party of course likes to push the idea of linguistic uniformity through Mandarin, and according to them that's what Chinese means. But since that isn't the definition our article is using, I don't know why our templates should use it. I also think that unless we move Mandarin Chinese to Chinese language, we should not call |lang=zh-cn "Chinese" either, but rather "Mandarin Chinese". Only plain |lang=zh should be "Chinese", and its use should be treated as an error, in my opinion, and we should consider raising one, maybe only putting it in an error category. Perhaps, Category:Articles with CS1 sources in unspecified language in Chinese language family.

Psiĥedelisto (talk • contribs) ^{please always ping!} 01:30, 19 July 2020 (UTC)

@Psiĥedelisto: Mandarin, Wu and Yue Chinese (the latter of which Cantonese is considered a dialect of) do have own ISO 639 codes: cmn, wuu and yue, respectively. Several other languages in the massive tent of written Chinese have separate codes as well, as is demonstratable by following links from Varieties of Chinese and looking at the articles' infoboxes. That doesn't solve this problem entirely, but it's incorrect to say that zh is the only available code. Glades12 (talk) 19:12, 19 July 2020 (UTC)

@Glades12: Thanks for clarifying this, I didn't know. It seems our template doesn't understand |lang=cmn, though. It does understand |lang=yue (outputting Cantonese) and |lang=wuu (outputting Wu Chinese). I guess why I say ISO dropped the ball is that I don't think |lang=zh should exist—no other broad language families have such codes, and it causes a lot of lazy mislabeling under the broad identifier rather than the narrow. What do you think about my idea to add a hidden error category? Now that you've taught me about |lang=cmn, I suppose we need that really: |lang=cmn-TW (Taiwanese Mandarin), |lang=cmn-CN (Standard Chinese, or should it be Beijing dialect?), etc. Psiĥedelisto (talk • contribs) ^{please always ping!} 19:21, 19 July 2020 (UTC)

@Psiĥedelisto: I'm undecided on an error category; simply a maintenance one, like Category:CS1 maint: unrecognized language, would probably be better. (Maybe that's what you meant by "hidden".) Correct me if I'm wrong, but there are probably a lot of cases where it's impossible to determine which variety an author uses in speech, since the vast majority look (more or less) identical in writing. In such cases (but not always, of course), I don't think we should penalise editors for using zh. Also, no idea why the software doesn't recognise cmn; it's as valid a code as the others we've mentioned here. Glades12 (talk) 07:36, 20 July 2020 (UTC)

Indeed, if we're citing a written text you can't tell what the author's native variety is, because they're writing in Standard Chinese, for which |lang=zh and the short name "Chinese" are quite appropriate. Codes like cmn, yue, wuu, etc would be relevant if we were citing recordings. Kanguole 08:32, 20 July 2020 (UTC)

@Kanguole: Not necessarily. Hokkien is often written in Tâi-lô or Tn̂g-lâng-jī, both of which are obviously different to Standard Chinese and not mutually intelligible, even in writing. Not making zh an error category is fine with me, though, I see your point. Cantonese, also, is often not fully mutually intelligible even in writing, and Cantonese-only characters exist. A Western reader might also confuse Chữ Nôm for Chinese, when they are also not mutually intelligible. So, these codes do serve a purpose for written texts. I'm not familiar enough with Wu Chinese to comment on it. Psiĥedelisto (talk • contribs) ^{please always ping!} 20:36, 20 July 2020 (UTC)

I quite agree that written Hokkien and written Cantonese ought to be marked specifically. That is a different situation from Standard Chinese as written in Taiwan, Hong Kong, Singapore or China. For Chữ Nôm, people are currently using |lang=vi-Hani. Kanguole 21:06, 20 July 2020 (UTC)

@Kanguole: OK but this is not really true. I don't know why you're insisting on the name "Standard Chinese" and not "Mandarin Chinese". "Standard" seems to elevate Mandarin unnecessarily, while Hokkien is itself a standard in Taiwan and is used in academic writing and news alongside Taiwanese Mandarin. Taiwanese Mandarin is a flavor of Mandarin with its own words and standards and authority over it. (Ministry of Education (Taiwan); National Languages Committee). See also Taiwanese Mandarin § Same words, different meaning et seq. We should be calling all Mandarin Mandarin, using cmn in most places we're using zh now. Psiĥedelisto (talk • contribs) ^{please always ping!} 02:10, 21 July 2020 (UTC)

Perhaps it is unfair that Guóyǔ ('national language') was codified in the early 1930s based on the Beijing dialect and Mandarin-based vernacular writing, but that is the reality, and pointing it out is not unnecessary elevation. Certainly the varieties of Taiwan and the mainland have diverged somewhat over the last 75 years, moreso in speech than in writing, but they are still the same language. Kanguole 07:43, 21 July 2020 (UTC)

Nothing against changing zh-tw to "Taiwanese Mandarin", adding cmn and putting zh into a tracking category, but since most Western readers won't be able to determine something more specific than zh when some text is advertised in the real world just as "Chinese" (which is often the case), it shouldn't be an error category as this would keep people from adding the information at all, and it is useful for formating purposes even while unspecific. --Matthiaspaul (talk) 17:59, 20 July 2020 (UTC)

I very much support this method; as someone who began editign the Taiwanese Mandarin article in earnest in May, and has big plans for it (not that none of us doesn't have big plans for a dozen separate articles at a time, but y'know), I have been trying to anticipate handling the number of pretty common tone/pronunciation differences as they pop up in the body of the article. A zh-tw would be more accurate and useful in circumstances where the ambiguity of "Chinese" might mislead a Mandarin learner to assume a Mainland utterance is grammatical in Taiwan or vice-versa, when it is not (that has happened to me before). WhinyTheYounger (talk) 02:22, 21 July 2020 (UTC)

others prerequisites

In the Usage section is a table of parameters with some having prerequisites. It notes that the others parameter has a prerequisite of title, but it fails to note that the others parameter also has a prerequisite of last or author. I would fix it myself but I can't figure out where this is stored. —Anomalocaris (talk) 07:12, 22 July 2020 (UTC)

|others= doesn't have that prerequisite, and shouldn't anyway, because some sources (such as signs and very short magazine articles) only credit certain roles such as illustrators or editors. In cases like those, the templates should absolutely not force editors to choose between guessing who wrote the material and not including any names at all. Glades12 (talk) 11:09, 22 July 2020 (UTC)

It indeed has (essentially) that prerequisite as indicated by the existence of Category:CS1 maint: others. --Izno (talk) 13:03, 22 July 2020 (UTC)

I didn't know about that tracking category, but my point (that not all sources include their authors) still stands. Glades12 (talk) 14:35, 22 July 2020 (UTC)

Sorry. I was commenting on Template:Cite AV media notes, and from that page I clicked "talk" and I assumed I was on Template talk:Cite AV media notes, but I'm actually here at Help talk:Citation Style 1. My comments relate to Template:Cite AV media notes, which really does have a table as I described. Glades12: Izno is correct. Using any of these templates with the others parameter without last or author parameter causes the article to be automatically placed in Category:CS1 maint: others, which is an error category. In the case of Template:Cite AV media notes, the intent is that the authoring parameter is for the author of the notes, not for the composer or the lyricist or librettist or musical performer. If someone does put the performing musician's name in the authoring parameter, the display looks OK but it's supplying incorrect metadata. A lot of phonograph records, cassettes and CDs come with media notes with no apparent author. For example, I'm looking now at the notes that came with a CBS Records 1987 CD, Mozart: The Flute Quartets (Jean-Pierre Rampal, Flute; Isaac Stern, Violin; Salvatore Accardo, Viola/Alto; Mstislav Rostroprovich, Cello). The notes are in English, German and French, without author, although the German does list the Übersetzung (translator). Most other record and CD notes also don't have a listed author. A page using the others parameter (for the four musicians, for example) would be placed in the error Category:CS1 maint: others. This is a mess that should be fixed somehow. —Anomalocaris (talk) 06:22, 23 July 2020 (UTC)

"Using any of these templates with the others parameter without last or author parameter causes the article to be automatically placed in Category:CS1 maint: others, which is an error category." Nope. There's a distinction between CS1 tracking categories (prefixed with "CS1 maint") and error categories (prefixed with "CS1 errors"). Pages in the tracking categories may need to be edited for possible errors, while ones in the error categories certainly contain errors and should be edited in any case. Also, there was literally no way to know that you were talking about Cite AV media notes specifically; everything here implied that you were talking about all CS1 templates. Glades12 (talk) 06:39, 23 July 2020 (UTC)

|title-link= prerequisite at {{cite AV media notes}} was misplaced; fixed now.

—Trappist the monk (talk) 09:59, 23 July 2020 (UTC)

Glades12: You're right, there was no way to know I was talking about Template:Cite AV media notes. As I explained, when I made my initial post here, I didn't realize I wasn't on Template talk:Cite AV media notes. I apologize for failing to note in my initial posting that I was talking about Template:Cite AV media notes. That's why I noted it as soon as I realized the issue.

Trappist the monk: You have removed all prerequisites from the others parameter, was this intentional?

Anyone interested: The subcategories of Category:CS1 maintenance may not be error categories, but many if not most of them seem to be de facto error categories. For example, Category:CS1 maint: archived copy as title‎ and Category:CS1 maint: ASIN uses ISBN‎ are de facto error categories. I don't believe I am the only Wikipedian who has assumed that if an a page is listed in any subcategory of Category:CS1 maintenance, it should be edited to coax the page out of the subcategory. I know there are Wikipedians who comb through these categories and fix things; within the past week or so, Category:CS1 maint: unrecognized language‎ had about 350 articles, and over the course of about two days, I saw the number drop to 42, and I checked the history of some articles that had been in the category and I saw that one or more editors were fixing the language parameter. Let us compare the error category Category:CS1 errors: URL–wikilink conflict to the non-error category Category:CS1 maint: others.

• Category:CS1 errors: URL–wikilink conflict: This is a tracking category for CS1 citations that have wikilinks embedded in |title=, |chapter=, |article=, or |booktitle=, etc. while also specifying a URL for that element.
• Category:CS1 maint: others: This is a tracking category for CS1 citations that use |others= without also using |author= or |editor= or any of their aliases. |others= is provided to record other (secondary) contributors to the cited source. Articles are listed in this category when Module:Citation/CS1 identifies a template that does not identify primary contributors....

So both categories claim to be tracking categories, but to the Wikipedia priesthood, one of them is an an error category and the other is not. I have been editing Wikipedia for over 15 years and I am in the top 1,300 Wikipedians by number of edits, but despite my status as a "senior editor", this distinction is becoming apparent to me only now. If Category:CS1 maintenance is not a collection of error category pages, I believe it and its member pages should me be modified to make the distinction clearer. Perhaps something like, "This is not an error category and pages listed listed here do not necessarily need to be 'fixed'." But we need to be careful, since any page in, for example, Category:CS1 maint: archived copy as title‎, does need to be fixed. —Anomalocaris (talk) 22:00, 23 July 2020 (UTC)

The entire citation area is a mess and has been for the last three or four years. A small group of people are making decisions that massively impact the functioning of the project and they're doing so in an out-of-the-way enclave. It has become so confusing anf frustrating that, frankly, I am on the verge of giving up. - Sitush (talk) 22:19, 23 July 2020 (UTC)

My intent was realignment of |title-link= with its prerequisite |title= because, as it was, it was so obviously wrong and because, at the time, I had other stuff on my mind. The incorrect alignment has been in place since I made this edit – I guess it wasn't so obvious at the time... Unless someone beats me to it, I'll add |lastn=, |authorn=, |editor-lastn=, and |editorn= as prerequisites for |author= (probably tomorrow).

—Trappist the monk (talk) 00:04, 24 July 2020 (UTC)

duplicate error message

Cite book comparison

Wikitext	{{cite book|author-link=[[Celia Fiennes]]|first=Celia|last=Fiennes|location=London|publisher=Leadenhall Press|title=Through England on a Side Saddle in the Time of William and Mary|year=1888}}
Live	Fiennes, Celia (1888). Through England on a Side Saddle in the Time of William and Mary. London: Leadenhall Press. {{cite book}}: Check |author-link= value (help)
Sandbox	Fiennes, Celia (1888). Through England on a Side Saddle in the Time of William and Mary. London: Leadenhall Press. {{cite book}}: Check |author-link= value (help)

fixed in the sandbox;

—Trappist the monk (talk) 12:53, 22 July 2020 (UTC)

Rendering page number in book-length computer files

We need a way to handle page numbers in computer files and eBooks that are not absolute page numbers. Google has their own way of identifying such pages, and we should decide how we want to handle it in the general case, Google or otherwise. I'm hoping for a discussion to see, in the first instance, whether there's interest in establishing a conventional method of doing this (achievable via /doc changes alone), or if something more robust is required.

Checked the archives, and didn't see anything about this; please add a link if I missed something. More and more, books are being "printed" online only, and exist purely as digital files. Or as both, but the accessible one is digital. Especially in the former case, there may not be an absolute "page number", depending on the format (i.e., not pdf or other fixed format) and on the rendering engine. In particular, Google will render these without visible page numbers in their page view mode. They do have an identifier they use in their url to distinguish the two cases. As near as I can determine from generalizing from a few dozen examples, the url param |pg= is used for both cases, but the value differs depending on the source; for example: |pg=PA35 for printed, absolute page number visible in printed version, and |pg=PT35 for a page number on a digital resource. Note that in book search results, the Google result snippet will be slightly different: the boxed contextual snippet will say found inside – page 35 in the latter case, and found inside in the former.

For starters, I think this could be a doc-only change, by way of some additional text at the section on page, recommending what to do in this case, without any need for software changes. For example, something like:

For computer files where no fixed page number is present, code the page number a |page=X99, where the 'X' prefix is replaced by an identifier ('G' for Google books, 'I' for internet archive, ...) and the '99' represents the page identifier given by that display provider. The following table provides the identifiers for some common eBook providers: <table>...

That's just a first cut, and I'm sure I failed to consider lots of things. But the point is, I think we can initiate something useful without a software change, which would be a lot easier to get going, n'est-ce pas?

An example of using a conventional approach, as proposed

In this example from Google books, page numbers are absent in both the results snippet and the "view inside" pages, but a value appears in the url. By June, the different branches of Free France, led by de Gaulle out of London, and by Giraud out of Algeria, merged into one, creating the French Committee of National Liberation.<ref name="Davis-2018">{{cite book |lang=en |last=Brunet |first=Luc-Andre |editor-last1=Davis |editor-first1=Muriam Haleh |editor-last2=Serres |editor-first2=Thomas |title=North Africa and the Making of Europe: Governance, Institutions and Culture |chapter=1. The Role of Algeria in Debates over Post-War Europe within the French Resistance |url=https://books.google.com/books?id=tP5DDwAAQBAJ&pg=PT35 |accessdate=23 July 2020 |date=22 February 2018 |publisher=Bloomsbury |isbn=978-1-350-02184-6 |oclc=1037916970 |type=computer file |page=G35–36}}</ref>

Going forward, maybe we do want more control of this, so maybe there's a new param to explain who the rendering engine is:(e.g., |epager=Google, |epager=Internet Archive, etc.). You would think that the value of the |url= field would be enough to imply the latter, but in the real world, the multiplicity of CS1 params not infrequently don't all remain in sync, so I wouldn't trust that method; you'd end up with page numbers corresponding to some mystery provider. That might even be a reason to keep the original suggestion (i.e., use |pg=G35 for Google efile) because the page number and the method are kept together in one param value. Your thoughts appreciated. Mathglot (talk) 01:10, 24 July 2020 (UTC)

I recommend citing the chapter, and if you have the energy, provide a quotation supporting the claim in the article. We have never required page numbers, and some sources, like web sites, have never had page numbers. I don't think it should be our job to make up nonexistent location designations for documents that do not have them. – Jonesey95 (talk) 03:15, 24 July 2020 (UTC)

There's a similar issue in scanned documents. There is a page within the PDF and there is the number that was printed on the original physical page; those are often different. Shmuel (Seymour J.) Metz Username:Chatul (talk) 04:29, 24 July 2020 (UTC)

Page numbers are used to help a reader discover the information that supports the wikitext. They are not necessarily pertinent in digital formats, which may use electronic bookmarks or other similar tagging as locators. Assuming that one can link that exact location (it is increasingly possible), the point of pagination is moot: the reader will be immediately directed to the verifying information. What may perhaps become useful in the future is better handling of such locators. Right now the only such specific locator used in CS1/CS2 is |chapter-url= and its aliases. Other than that, I don't think any special system is warranted. In scanned works, what is actually cited is the scan image, not its source. Different considerations apply, imo. 98.0.246.242 (talk) 01:37, 25 July 2020 (UTC)

Template:Cite book needs additional terms

"Template:Cite Book" needs terms for all of the MARC 21 fields, especially total pages, size, etc.71.230.16.111 (talk) 06:47, 24 July 2020 (UTC)

Why? The fields you mention are intended for cataloguing books. We are citing them, not cataloguing them. This information does not usually go into citations according to most commonly-used academic citation standards. Making fields for this information will just encourage people to fill them in under circumstances where they would be better off omitted. —David Eppstein (talk) 06:51, 24 July 2020 (UTC)

Agreed. It's actually confusing and detrimental to citations to add claptrap like this, especially total page count, since it's often mistaken for the pages being cited for the information the citation pertains to. And we just have utterly no use for something like "|size=quarto". Yeesh. Wikipedia is not anything like WorldCat.  — SMcCandlish ☏ ¢ 😼  13:00, 4 August 2020 (UTC)

"References" are not just citations. A "citation" only needs to identify a known reference; we need to describe the reference, for readers not able to just go over to a shelf and pick it up, and to convince the reader that it is in fact a valid reference for the subject of the article. Thus the academic standard does not apply; whether something is published as manuscript, paperback or hardcover does matter. Should there be a different template, "identify book" instead of "cite book"?71.230.16.111 (talk) —Preceding undated comment added 06:11, 7 August 2020 (UTC)

Fulltext access fields

I'm sure the way in which access is indicated in citations has been discussed extensively by the regulars here, and I apologize for my ignorance of your past discussions. Is there a field to indicate that a book is public-domain (in this case, a 2018 work of US government) and the fulltext is freely available online? This will make it obvious that it is easy to verify the cited statement, and increase the likelihood of editors and readers actually doing it. I understand that such access parameters are already common on journal article templates.

An additional winkle here: many medical journal articles are currently freely accessible due to a special action taken by publishers for the COVID-19 pandemic. At some future date to be decided upon by the publishers, they will put the articles back behind paywalls. Public licenses are permanent, but free access may be revoked. I'm not sure how widespread this is, but we might end up with a lot of incorrect information on access. HLHJ (talk) 17:29, 8 August 2020 (UTC)

Does Help:Citation Style 1 § Registration or subscription required answer your question?

—Trappist the monk (talk) 17:36, 8 August 2020 (UTC)

|subject-link= and |subject-mask=

The other day, I noticed that we don't have |subject-mask= or any of its enumerated forms. I have added |subject-mask=, |subjectn-mask=, and |subject-maskn=

{{cite interview/new |title=Title |subject=Abraham Lincoln |subject-mask=2}}

——. "Title" (Interview).

{{cite interview/new |title=Title |subject=Abraham Lincoln |subject1-mask=2}}

——. "Title" (Interview).

{{cite interview/new |title=Title |subject=Abraham Lincoln |subject-mask1=2}}

——. "Title" (Interview).

The |subject= and |interviewer= arrays of parameters are used primarily in {{cite interview}}. Because we don't have non-hyphenated forms of the |interviewer= parameters and because the preferred form for parameter names is hyphenated, I have deprecated |subjectlink=, |subjectlinkn=, and |subjectnlink=. Here are some simple searches that indicate usage of these parameters:

• |subjectlink=: ~860 hits
• |subjectlinkn=: ~110 hits
• |subjectnlink=: times out
  • constrained to {{cite interview}}: |subjectnlink= 0 hits

—Trappist the monk (talk) 22:48, 26 July 2020 (UTC)

|website= parameter

​How should editors use the |website= parameter? I'm talking about this edit and similar ones. The examples at Template:Cite web#Examples use the website name (in this case, Parties and Elections in Europe), but the automatic citation-creating tool uses the url of the website (in this case, www.parties-and-elections.eu). However, I also tried to use the auto-citation tool with a Wikipedia article as the url to test, and it put Wikipedia in the website parameter. Is this a mistake in the auto-citation tool? Thanks, Ezhao02 (talk) 15:37, 27 July 2020 (UTC)

The tool in question uses a system which has pre-configured settings for some websites (that you could get involved with if interested) and defaults to a certain setting. For e.g. Wikipedia, that setting indicates that Wikipedia is the website. For e.g. Parties and Elections, that's the default so it adds the website main URL. You should prefer the name if available and provide the URL if unavailable. --Izno (talk) 15:55, 27 July 2020 (UTC)

@Izno: To be clear, are you saying that both www.parties-and-elections.eu and Parties and Elections in Europe should be listed, or are you talking about the url of the web page (in this case, http://www.parties-and-elections.eu/basquecountry.html) being put in the url parameter? Is either the way I formatted it (prior to Impru20's edit linked above) or the way Impru20 formatted it considered "right" or "wrong"? Thanks, Ezhao02 (talk) 15:59, 27 July 2020 (UTC)

both www.parties-and-elections.eu and Parties and Elections in Europe should be listed No, do pick one or the other. Our guidance at Help:CS1#Work and publisher is currently the following:

On websites, in most cases "work" is the name of the website (as usually given in the logo/banner area of the site, and/or appearing in the <title> of the homepage, which may appear as the page title in your browser tab, depending on browser). Do not append ".com" or the like if the site's actual title does not include it (thus |work=Salon, not Salon.com). If no clear title can be identified, or the title explicitly is the domain name, then use the site's domain name. Do not falsify the work's name by adding descriptive verbiage like "Website of [Publisher]" or "[Publisher]'s Homepage". Capitalize for reading clarity, and omit "www.", e.g. convert "www.veterinaryresourcesuk.com" to "VeterinaryResourcesUK.com".

Does that help you decide? --Izno (talk) 16:02, 27 July 2020 (UTC)

Yes, thank you! Ezhao02 (talk) 16:18, 27 July 2020 (UTC)

@Izno: I still have another question about that page. How would you decide if "the title explicitly is the domain name"? Ezhao02 (talk) 16:23, 27 July 2020 (UTC)

Basic duck test. Does it look like the same duck? Yes. Then it's probably the same duck. --Izno (talk) 16:28, 27 July 2020 (UTC)

Thank you for all your help! Ezhao02 (talk) 16:31, 27 July 2020 (UTC)

Just to clarify further about "both www.parties-and-elections.eu and Parties and Elections in Europe should be listed ...?" Izno and the cited documentation are correct in answering "No." Either is okay; neither are an error, but both is an error. If the work has a proper title, then use that, but the domain name will suffice in absence of one, or if you don't have time to go look. It's okay that automated tools use the domain name, and it's also okay to cleanup up after them and change the auto-cites to use real titles. For a work without a title or whose title is literally its domain name, it's best to shave off the "www." if the site can be reached without it (most of them can, but it's worth testing; I ran into one only two days ago that 404'ed without the "www." prepended to it).  — SMcCandlish ☏ ¢ 😼  12:56, 4 August 2020 (UTC)

Support contribution/section in cite journal and others.

• {{Cite journal | first1 = Shana | last1 = Kusin | first2 = Teddy | last2 = Angert | first3 = Katie | last3 = von Derau | first4 = B. Zane | last4 = Horowitz | first5 = Sandy | last5 = Giffin | name-list-format = vanc | title= 2012 Annual Meeting of the North American Congress of Clinical Toxicology (NACCT) October 1–6, 2012 las Vegas, NV, USA | volume = 50 | issue = 7 | pages = 574–720 | url = http://www.ohsu.edu/emergency/about/news/2012/nacct/posters/squash.pdf | contribution= 189. Toxic Squash Syndrome: A case series of diarrheal illness following ingestion of bitter squash, 1999-2011 |journal=Clinical Toxicology | doi = 10.3109/15563650.2012.700015| year = 2012 }}

Gives

• Kusin, Shana; Angert, Teddy; von Derau, Katie; Horowitz, B. Zane; Giffin, Sandy (2012). "2012 Annual Meeting of the North American Congress of Clinical Toxicology (NACCT) October 1–6, 2012 las Vegas, NV, USA" (PDF). Clinical Toxicology. 50 (7): 574–720. doi:10.3109/15563650.2012.700015. {{cite journal}}: |contribution= ignored (help); Unknown parameter |name-list-format= ignored (|name-list-style= suggested) (help)

It should give something better, like

• Kusin S, Angert T, von Derau K, Horowitz BZ, Giffin S (2012). "189. Toxic Squash Syndrome: A case series of diarrheal illness following ingestion of bitter squash, 1999-2011" in "2012 Annual Meeting of the North American Congress of Clinical Toxicology (NACCT) October 1–6, 2012 las Vegas, NV, USA". Clinical Toxicology. 50 (7): 574–720. doi:10.3109/15563650.2012.700015.

Or maybe

• Kusin S, Angert T, von Derau K, Horowitz BZ, Giffin S (2012). "2012 Annual Meeting of the North American Congress of Clinical Toxicology (NACCT) October 1–6, 2012 las Vegas, NV, USA: 189. Toxic Squash Syndrome: A case series of diarrheal illness following ingestion of bitter squash, 1999-2011". Clinical Toxicology. 50 (7): 574–720. doi:10.3109/15563650.2012.700015.

Headbomb {t · c · p · b} 18:31, 27 July 2020 (UTC)

Pinging @David Eppstein: since you have a lot of experience there with math citations to sections of journal articles/books. Headbomb {t · c · p · b} 18:33, 27 July 2020 (UTC)

Use |department=:

• {{Cite journal | first1 = Shana | last1 = Kusin | first2 = Teddy | last2 = Angert | first3 = Katie | last3 = von Derau | first4 = B. Zane | last4 = Horowitz | first5 = Sandy | last5 = Giffin | name-list-format = vanc | department= 2012 Annual Meeting of the North American Congress of Clinical Toxicology (NACCT) October 1–6, 2012 las Vegas, NV, USA | volume = 50 | issue = 7 | pages = 574–720 | url = http://www.ohsu.edu/emergency/about/news/2012/nacct/posters/squash.pdf | title= 189. Toxic Squash Syndrome: A case series of diarrheal illness following ingestion of bitter squash, 1999-2011 |journal=Clinical Toxicology | doi = 10.3109/15563650.2012.700015| year = 2012 }}

• Kusin, Shana; Angert, Teddy; von Derau, Katie; Horowitz, B. Zane; Giffin, Sandy (2012). "189. Toxic Squash Syndrome: A case series of diarrheal illness following ingestion of bitter squash, 1999-2011" (PDF). 2012 Annual Meeting of the North American Congress of Clinical Toxicology (NACCT) October 1–6, 2012 las Vegas, NV, USA. Clinical Toxicology. 50 (7): 574–720. doi:10.3109/15563650.2012.700015. {{cite journal}}: Unknown parameter |name-list-format= ignored (|name-list-style= suggested) (help)

—David Eppstein (talk) 18:40, 27 July 2020 (UTC)

That is very hacky. Headbomb {t · c · p · b} 17:16, 30 July 2020 (UTC)

It is, as far as I know, the only way to get three levels of titling (title/department/journal) within a journal cite. The other way is to use chapter/title/series or contribution/title/series but then you would have to call Clin.Tox. a series rather than a journal, and the volume/issue formatting would be for a book not a journal, so that's worse. The department is not included in the COinS metadata (and in particular is not coded wrongly in COinS as a department rather than as the title of a special issue). —David Eppstein (talk) 17:35, 30 July 2020 (UTC)

Please add param error checking for interwikis lacking leading colon

Hello, can you please add some error-check code and emit a big, red error message, if an editor codes a "link" param (author, title, translator, editor, contributor are the ones I'm aware of) to a sister wikipedia like fr-wiki without a leading colon on the lang code? What happens in this case, is various kinds of strange behavior, including dropping the field (author, say) and emitting only a semicolon or other punctuation; but worse is that the first of these, whatever it is, is picked up as an interwiki, and the entire article is linked under that language name in the left sidebar. Same thing occurs for fields with permissible wiki-linking, such as publisher. Beyond that, it screws up article links at Wikidata, as soon as their bot sees it.

You can see an example of the faulty behavior in revision 969921029 of André Diethelm. Go to the language links in the left sidebar, and notice that the Hebrew and Arabic links point to the correct articles (Google translate does a sufficient job of transliterating the titles to confirm they are correct if you need it) but the French one is wrong, and links to fr:Éditions Philippe Rey. This happens to be the publisher of one of the sources listed in the #Further reading section, namely "Lambert (2010)", but you won't see it on the rendered page, because it's been snatched by the wikimedia code and interpreted as the target of the "Francais" link in the language sidebar; all you will see in the Lambert citaton is extra punctuation between "Paris" and the ISBN. To actually see the linked publisher with the missing colon, you have to view the wikicode of that revision.

You can diff that version with current (diff), to see that the only change was to add one colon in the Lambert publisher field. Notice that the Francais link in the sidebar is now correct. (Same thing happens if you forget the leading colon in authorlink, or any of the other *-link fields.) These should be flagged as errors, because they will escape the notice of many users. Also, the bots at Wikidata are rapid, and although I noticed and fixed the problem within minutes, I was too late, and the Wikidata bot had already linked my French Resistance guy at en:André Diethelm to the French publisher fr:Éditions Philippe Rey via item d:Q3579477 ("Éditions Philippe Rey"; diff). I had to both unlink that connection, then go relink Diethelm to d:Q2847654 instead (same one that has the Hebrew and Arabic articles as well).

This is too much to ask most users to do. A big, fat, red error for interwiki links that lack a leading colon would be a big help. Thanks, Mathglot (talk) 06:04, 28 July 2020 (UTC)

I added a request for such error handling at Help talk:CS1 errors. Thanks, Mathglot (talk) 09:55, 3 August 2020 (UTC)

That discussion closed. One discussion in one place.

—Trappist the monk (talk) 20:06, 4 August 2020 (UTC)

The |publisher=[[fr:éditions Philippe Rey|Philippe Rey]] is not a |<param-link>= problem; somewhat related, but not the same.

I have hacked the module sandbox to catch |<param-link>= where the assigned value begins with a known local inter-wiki prefix without a leading colon. There are a couple of exceptions. The w: and wikipedia: prefixes do not require leading colon. The s: and wikisource: prefixes have special meaning because cs1|2 creates urls from these inter-wiki links so that it can add the wikisource icon. From the example template, tweaked so that the author wikilink is missing the leading colon:

{{Cite book/new |first1=Gilles |last1=Lambert |author1link=fr:Gilles Lambert |title=Title}}

Lambert, Gilles. Title. {{cite book}}: Check |author1link= value (help)

Inter-wiki links are apparently namespace sensitive. The above example shows a linked author name and there is no link under languages to the fr.wiki. When I copied the above example to a random article in en.wiki main space and previewed, the linked author name is missing and fr.wiki is listed in the languages list, linked to Gilles Lambert.

The |publisher= inter-wiki link problem must be handled differently because for the most part, any cs1|2 parameter may be wiki-linked (despite documentation to the contrary). For all parameters in a cs1|2 template that are not |<param-link>= parameters, inter-wiki links must begin with [[<prefix>: where <prefix> is a known local inter-wiki prefix, s:, w: and their long-forms again excepted. As each parameter name is validated, the cs1|2 sandbox looks for this pattern and emits and error message when detected:

{{Cite book/new |title=[[fr:Title]]}}

fr:Title. {{cite book}}: Check |title= value (help)

In main space, the title from the above example is missing and fr.wiki is listed in the languages list, linked to Title.

—Trappist the monk (talk) 20:06, 4 August 2020 (UTC)

There are other kinds of interwiki links that should (could?) be supported such as d:; interlanguage links are the ones that are problematic only. I do not understand why w: and wikipedia: (Wikipedia here is presumably its use as an interlanguage link as in wikipedia:Wikipedia:BRD?) are supported in this way. --Izno (talk) 21:26, 4 August 2020 (UTC)

I included the wikipedia: in the exclusion list because it is listed at Help:Interwiki linking as the long form of w:. But, now that you point it out, wikipedia: the prefix makes no sense because it conflicts with wikipedia: the namespace. No doubt the exclusion list might include b: (wikibooks), c: (commons), d: (wikidata), m: (meta), n: (wikinews), q: (wikiquote), v: (wikiversity).

—Trappist the monk (talk) 22:11, 4 August 2020 (UTC)

wikipedia: does not conflict on other wikis as a thought. --Izno (talk) 22:55, 4 August 2020 (UTC)

[[wikipedia:]] on meta: links to the en.wiki main page. As a guess, I would say that wikipedia:, the name space, is common on encyclopedia wikis but not on other types of wiki.

Because [[w:]] appears to be a prefix that always links to en.wiki, I'm wondering if that prefix should continue to be excluded from the error check. Here at en.wiki, the w: prefix gets you to the article in the same way that the en: prefix or no prefix does:

[[w:Abraham Lincoln]]

[[en:Abraham Lincoln]]

[[Abraham Lincoln]]

At other wikipedias, the w: prefix links to the en.wiki article but isn't otherwise treated as a inter-language inter-wiki link.

—Trappist the monk (talk) 12:53, 5 August 2020 (UTC)

Regarding, "Inter-wiki links are apparently namespace sensitive", yes; see Help:Interlanguage links#Method. And that's great that you were able to mock up something in the sandbox for the *-link case so quickly. I realize the other case (publisher, and other fields) is different, and is complicated by the legit shortcut codes that link to sister projects. Did anyone mention wikt:, species:, v:, voy:? See also, m:Help:Interwiki linking. Mathglot (talk) 09:35, 5 August 2020 (UTC) updated 10:21, 5 August 2020 (UTC)

I mentioned v: earlier.

At present cs1|2 tests what it thinks is a prefix against the table of known inter-wiki prefixes returned from mw.site.interwikiMap ('local'). I am minded to change that and instead use the list of languages returned by mw.language.fetchLanguageNames ('<local wiki lang code>', 'all') if I can show that all of the language codes in that list are also found in the inter-wiki map list.

—Trappist the monk (talk) 12:53, 5 August 2020 (UTC)

I have changed the code so that ~/Configuration/sandbox creates a list of language prefixes from both mw.site.interwikiMap ('local') and mw.language.fetchLanguageNames ('<local wiki lang code>', 'all'). Prefixes in mw.site.interwikiMap ('local') must match a language code in mw.language.fetchLanguageNames ('<local wiki lang code>', 'all') to be added to the local list. There are seven 'language-like' codes in mw.site.interwikiMap ('local') that 'redirect' to another-language wiki but these codes do not contribute to the inter-wiki language list:

cmn: → Mandarin Chinese (ISO 639-3 code); redirects to zh.wikipedia.org

cz: → Czech (ISO 3166 country code); redirects to cs.wikipedia.org

dk: → Danish (ISO 3166 country code); redirects to da.wikipedia.org

epo: → Esperanto (ISO 639-3 code); redirects to eo.wikipedia.org

jp: → Japanese (ISO 3166 country code); redirects to ja.wikipedia.org

minnan: → invalid IETF language tag; redirects to zh-min-nan.wikipedia.org

zh-cfr: → invalid IETF language tag; redirects to zh-min-nan.wikipedia.org

These are not included in the local prefix list.

—Trappist the monk (talk) 15:56, 5 August 2020 (UTC)

Can't remember anymore where I saw it, but there's a case where one type of Norwegian language code—whether no (Norwegian, in general) or nn (Nynorsk), I can't remember —is odd man out, wrt to a giant list of WP codes that generally match the ccTLD codes, except for that one case. (There's also Bokmal, nb, but it wasn't that.) This could be a red herring, but just wanted to recall it, in case it's relevant here. Mathglot (talk) 03:40, 6 August 2020 (UTC)

I'm not sure I understand what it is that you are saying. [[nn:]] links to the Norsk nynorsk nn.wiki, both of [[nb:]] and [[no:]] link to the Norsk bokmål nb.wiki. I do remember that for a time, language code no was not supported by the {{#language:}} magic word. That has since been fixed:

{{#language:nb|en}} → Norwegian Bokmål

{{#language:nn|en}} → Norwegian Nynorsk

{{#language:no|en}} → Norwegian

—Trappist the monk (talk) 10:30, 6 August 2020 (UTC)

Found it, and it doesn't affect this issue, because it's about translation, not linking. If you're curious, it's used at Template:Expand Norwegian (Nynorsk); note the diff between |langcode= and |googlelangcode=, not found in any other Expand language template (such as Template:Expand Norwegian). The explanation at this /doc page references Nynorsk. So, red herring, as far as we are concerned here. Mathglot (talk) 04:28, 7 August 2020 (UTC)

Also, while I'm thinking of it: despite the emission of a CS1 error in preview mode, the editor can still override and save anyway, as shown in your sandbox example above. This is fine, for the *-link case, but not so great for the publisher=[[fr:Seuil]] case, because if that link is saved like that, the Wikidata bot will likely do something strange. I don't know if this is beyond the scope of this template, but, for example, could you stop the user from saving it in that form, either by stripping out the brackets, or supplying the preceding colon? If not, maybe in that case we'd need to see about an edit filter to trap it, but I suppose that would be a discussion for somewhere else. Mathglot (talk) 09:50, 5 August 2020 (UTC)

Hm, maybe not so fine in the *-link case, either; they can both lead to downstream problems of rendering and knock-on issues at Wikidata, if I'm not mistaken. Mathglot (talk) 03:47, 6 August 2020 (UTC)

Tweaks to the code:

For |<param-link>= parameters that fail the various link tests (has a url, has wikilink markup, has inter-wiki prefix without leading colon) the code simply unsets the parameter and declares the error:

{{Cite book/new |title=Title |title-link=//example.com}} → Title. {{cite book}}: Check |title-link= value (help); External link in |title-link= (help)

{{Cite book/new |title=Title |title-link=[[Title]]}} → Title. {{cite book}}: Check |title-link= value (help)

{{Cite book/new |title=Title |title-link=nv:Title}} → Title. {{cite book}}: Check |title-link= value (help)

For |<param>=[[<prefix>:<value> the code extracts the label portion of the wikilink:

{{Cite book/new |title=[[es:Title]]}} → es:Title. {{cite book}}: Check |title= value (help)

{{Cite book/new |title=[[es:Title|Title]]}} → Title. {{cite book}}: Check |title= value (help)

These tweaks prevent the addition of extraneous inter-wiki links in the languages list.

That languages inter-wiki list at left must have a proper name. What is that name?

—Trappist the monk (talk) 13:31, 6 August 2020 (UTC)

Usually just inter-language list or inter-wiki list. The latter can be ambiguous since the introduction of the inter-project list. --Izno (talk) 14:33, 6 August 2020 (UTC)

I've always just called it the "language links sidebar" (or "language sidebar", on the second occasion) because I've never seen anyone misunderstand what that means, but I don't know if there's an official term.

Thanks so much for tweaks here and earlier; this is going to really improve things. I wonder how many "stealth errors" are out there—defining that as something that "seemed all right" until now, that suddenly will generate a CS1 error where it didn't before. Do we care? I imagine editors at the articles concerned will wonder why something turned red that never was before, but all they have to do is click the [help] to find out. This is great stuff, thanks Trappist. Mathglot (talk) 03:07, 7 August 2020 (UTC)

Proposal to support link decoration not only for links to Wikisource but also for other inter-wiki links

@Trappist the monk: Above you wrote: "The s: and wikisource: prefixes have special meaning because cs1|2 creates urls from these inter-wiki links so that it can add the wikisource icon." I didn't know that and thought it would be a special feature of {{cite wikisource}} rather than a general CS1/CS2 feature. However, trying it out I found that it works only with prefix "s:", not with ":s:":

{{cite book |title=Aard-vark |title-link=s:1911 Encyclopædia Britannica/Aard-vark |work=Encyclopædia Britannica |date=1911}}

Aard-vark . 1911. {{cite book}}: |work= ignored (help)

{{cite book |title=Aard-vark |title-link=:s:1911 Encyclopædia Britannica/Aard-vark |work=Encyclopædia Britannica |date=1911}}

Aard-vark. 1911. {{cite book}}: |work= ignored (help)

Shouldn't this work for both variants?

Actually, per the principle of least surprise, I think, it would be useful to add the same functionality also for links to other inter-wiki-links (iwl), but at least for links to the two most important ones, Wikidata (:d:/:wikidata:) and Commons (:c:/:commons:).

In the case of iwls, the corresponding project logos could be used as icons (as per Wikisource).

If we would extend the idea to inter-language-links (ill) as well and make it a general principle, we should not display iconized country flags, but instead just append a small text postfix composed of the prefix in square brackets (so, prefix :de: would result in [de] being appended to the link, similar to what links created by {{ill}} look like).

--Matthiaspaul (talk) 10:09, 9 August 2020 (UTC)

The s: and :s: prefixes work as they should work. When the {{cite wikisource}} parameter |noicon= is present and has a value (do not display the icon), {{cite wikisource/make link}} creates an inter-wiki link with the :s: prefix; else {{cite wikisource/make link}} creates an inter-wiki link with the s: prefix to show the icon.

—Trappist the monk (talk) 11:01, 9 August 2020 (UTC)

Okay, then it is a feature. But is it a useful feature to have this choice (as part of the link syntax being used, that is)? (This certainly could be controlled in different ways as well.)

I would think that in the majority of cases displaying the link decoration would be preferable - after all, we also cannot (at least not on user level) disable the external link arrow icon being displayed for external links. And in cases, where no icons should be displayed for some reason, it would be a template-wide rather than a link-specific setting (therefore having an option like |noicon= to control it).

Either way, if we have it for Wikisource, why not for Wikidata and Commons as well for reasons of consistency?

--Matthiaspaul (talk) 11:58, 9 August 2020 (UTC)

Citing third-party sources embedded in document

I'm trying to cite a newspaper source, for which the paper (or its website) doesn't exist anymore, but copies of the relevant article are embedded in a council ordinance. Can I do that, and if so, what would be the syntax?
The full ordinance is here, and I would like to reference page 106: "Success or failure?", Sean Robinson, The News Tribune. Frescard (talk) 14:23, 29 July 2020 (UTC)

If I understand correctly, you want to cite the newspaper article directly, as a sort of substitute stand-alone reference. If that is so, the answer is no, that is not correct. We must refer to sources as they are likely to be found. Here, the source is documentation of a local government ordinance. That is what should be cited. A location in the source is pertinent to the relevant wikitext. So the citation should be formatted exactly like that. Per your description, a reference to the particular news article cannot be sourced any other way. The answer was in the question. 172.254.241.58 (talk) 18:45, 30 July 2020 (UTC)

You could cite it as one document in a couple of ways (probably making use of |via=), but if you think it is helpful, you can instead cite both in one <ref> statement; something like {{cite news |newspaper=The News Tribune |first=Sean |last=Robinson |title=Success or failure? Probing Prometa}} as in {{cite web |url=https://online.co.pierce.wa.us/cfapps/council/model/otDocDownload.cfm?id=1448900&fileName=2007-81s%20final%20Ord%20file%201.pdf |department=Today in the Trib |website=Pierce County, Washington |title=Ordinance No. 2007-81s: etc. |p=106}}: Robinson, Sean. "Success or failure? Probing Prometa". The News Tribune. as in "Ordinance No. 2007-81s: etc" (PDF). Today in the Trib. Pierce County, Washington. p. 106. --Izno (talk) 02:18, 31 July 2020 (UTC)

The OP mentions that the original (the newspaper) is no longer in existence and cannot be found. If that is the case, citing the article directly is misleading and unverifiable. The use of |via= is I believe problematic: the source that includes the article is not an archive, repository, or other publisher. The article is included as supporting (incidental) documentation, as the OP described it, a source from a 3rd party. Tertiary sources should be referenced as such. 98.0.246.242 (talk) 02:16, 1 August 2020 (UTC)

He has stated a belief the web content cannot be accessed any longer directly. It may still be present in archives, digital or in fact physical. Giving the details of the document itself and stating that it can be found in the council minutes doesn't mislead whatsoever and may in fact aid someone to find it in one oc those other two contexts. Via is used to indicate a republisher. A republisher does not need to be one of those groups that you limited it to and in fact the council here is acting as a republisher of the original report. --Izno (talk) 03:00, 1 August 2020 (UTC)

WP:SWYGT. Also, archives (not self-archiving services) and legitimate republishers are supposed to satisfy legal/copyright requirements and are liable. In those cases, use of |via= is justified. But even that case does not apply here. The relevant text is a fragment of the containing source. It should be cited as such. 65.88.88.57 (talk) 14:16, 2 August 2020 (UTC)

auto-link bug fix

Cite journal comparison

Wikitext	{{cite journal|bibcode=1978BAMS...59..702B|doi-access=free|doi=10.1175/1520-0477(1978)059<0702:hbdstf>2.0.co;2|first=R. G.|issn=0003-0007|issue=6|journal=Bulletin of the American Meteorological Society|last=Barry|pages=702–705|title=H.-B. de Saussure: The First Mountain Meteorologist|volume=59|year=1978}}
Live	Barry, R. G. (1978). "H.-B. de Saussure: The First Mountain Meteorologist". Bulletin of the American Meteorological Society. 59 (6): 702–705. Bibcode:1978BAMS...59..702B. doi:10.1175/1520-0477(1978)059<0702:hbdstf>2.0.co;2. ISSN 0003-0007.
Sandbox	Barry, R. G. (1978). "H.-B. de Saussure: The First Mountain Meteorologist". Bulletin of the American Meteorological Society. 59 (6): 702–705. Bibcode:1978BAMS...59..702B. doi:10.1175/1520-0477(1978)059<0702:hbdstf>2.0.co;2. ISSN 0003-0007.

Fixed in the sandbox.

—Trappist the monk (talk) 16:19, 29 July 2020 (UTC)

Thank you! − Pintoch (talk) 20:04, 29 July 2020 (UTC)

Newspaper Article on Two Pages

What is the standard, if any, for citing a newspaper article that is available online, that is split onto two webpages. This occurs a lot with Newspaper.com sources. Generally speaking, the article is one source, with one title, author, date, etc. However, to assist with verifiability and to be able to archive each page of the article, I have been splitting it into two separate references, adding "Part 1" and "Part 2" to the title. See an example below:

• Dow, Bill (January 11, 2002). "Mann pioneer player in NFL: Part 1". Detroit Free Press (clipping). p. 1D. Archived from the original on November 22, 2019. Retrieved November 22, 2019 – via Newspapers.com.
• Dow, Bill (January 11, 2002). "Mann pioneer player in NFL: Part 2". Detroit Free Press (clipping). p. 8D. Archived from the original on November 22, 2019. Retrieved November 22, 2019 – via Newspapers.com.

This allows me to cite each sentence in an article with the relevant page of the news article. But it does create some confusion and adds additional sources in the reference list. Is there a better way to do this? Does (or could) {{Cite news}} support two urls and two archive-urls? Thanks for any insight. « Gonzo fan2007 (talk) @ 15:39, 3 August 2020 (UTC)

I'd combine the two thusly:

• Dow, Bill (January 11, 2002). "Mann pioneer player in NFL". Detroit Free Press. pp. 1D, 8D. Retrieved November 22, 2019 – via Newspapers.com.

I don't think an archive URL is needed with Newspapers.com. Even if the website ever shut down, you're ultimately citing the original newspaper via an accurate facsimile, and the URL is a courtesy, not the source itself, per se.

Also, I don't think |type=clipping is helpful because Newspapers.com will allow readers without an account to see the full page from which the clipping is drawn. Imzadi 1979 → 20:56, 3 August 2020 (UTC)

Add wayback-timestamp parameter

When an archive is added to a reference, the vast majority of the time it is just a Wayback Machine archive of the exact same URL. This bloats the source code of pages massively. It would be much simpler if a wayback-timestamp parameter was added, which would be set to the timestamp of the archive found in the page's URL. This was mentioned seven years ago here but the discussion had no conclusion. Example: |wayback-timestamp=20200721125421 in {{cite web|url=https://example.com/page|title=Example page|website=Example.com|date=2020-08-04|wayback-timestamp=20200721125421|archive-date=2020-07-21}} as opposed to the bloated {{cite web|url=https://example.com/page|title=Example page|website=Example.com|date=2020-08-04|archive-url=https://web.archive.org/web/20200721125421/https://example.com/page|archive-date=2020-07-21}}. Implementation: if waybackTimestamp then archiveUrl = 'https://web.archive.org/web/' + waybackTimestamp + '/' + url end.  Nixinova   T   C   05:15, 4 August 2020 (UTC)

You're not wrong, but the thing is, server space keeps getting cheaper and cheaper, and programmer (paid, or volunteer time) keeps getting more expensive and scarcer. If you had to prioritize this against stuff that's either broken and needs fixing, or enhancements that would provide desired new functionality, well, you see the problem... Mathglot (talk) 04:00, 6 August 2020 (UTC)

Smart substitution token to reduce redundancy among input parameters

I would propose a somewhat different implementation of this introducing a generic placeholder * into the syntax of the |archive-url= parameter. This would result in:

{{cite web |url=https://example.com/page |title=Example page |website=Example.com |date=2020-08-04 |archive-url=https://web.archive.org/web/20200721125421/* |archive-date=2020-07-21}}

The substitution would happen only if the |archive-url= contains exactly one * (not necessarily at the end of the link).

* is not normally used in urls (even less in archive links), but it can. In the case of archive.org, it is used as a wildcard for the timestamp, however, this isn't used in valid archive links (only internally by the template to help users select a specific snapshot). To not cause misinterpretation, this archive.org special syntax would have to be special-cased, so that if the * would be located inside the timestamp no substitution by |url= should happen.

My proposal is not as short as your proposal, but it is more flexible and could work also for a number of other archivers - and without having to add a new parameter or even a bunch of special |...-timestamp= ones.

The main reason, however, why I used * is another more generic feature proposal I planned to make for quite some while, where * would be a "smart" context-sensitive placeholder also supported in various other parameters:

It would substitute whatever is a sensible replacement string in the context of a particular parameter:

• For example, |access-date=* would be substituted with the value of |date= (if given, otherwise an error would be thrown instead of silently ignoring the parameter as what would happen for an empty parameter).
• |archive-date=* would be substituted with the value of |access-date= (likewise).
• |website=* would extract the domain name from |url=.
• |title-link=... * ... would implant the value of |title= at the position of the * (f.e. |title-link=* (song). Our |title-link=((...)) "take it as it is" ((syntax)) would allow for |title-link= to actually contain a * - in this case, the parameter substitution would be disabled. (TBD. What would be the best substitution if |trans-title= and/or |script-title= were used as well?)

Example: ... {title=Flying Circus |title-link=Monty Python's * (album) |date=...

• |author-linkn=... * .../|editor-linkn=... * ... etc. In the simple case, the substitute for * would be the value from |authorn=/|editorn=, the ((syntax)) would be supported as well. If the name is composed from multiple parameters such as |author-firstn=/|author-lastn=, the substitute would be the resulting string. This would allow for things like:

... |author-first=William |author-last=Shakespeare |author-link=* (author) |date=...

or

... |author-first=Otto |author-last=Sanders |author-link=:de:* (actor) |date=...

A single * would result in the composition of "<first> <last>" (because this is the most likely substring used in article titles), a doubled ** would result in "<last>, <first>" (comma, semicolon etc. depending on other template settings), a triple *** in only the "<last>" name. If someone would use ** or *** in |author-linkn= in conjunction with |authorn= rather than |author-firstn=/|author-lastn=, this would result in the contents taken from |authorn= as well (like a single *).

(TBD. At some point in the future we will probably support the full set of |trans-authorn=/|script-authorn= parameters. If they exist, the source values for the substitution should be derived from these parameters. The exact patterns are still TBD to maximize the utility value.)

• |author-maskn=... * ... etc. would support substitution as well. In this case, * would result in the composition of <first> and <last> according to the default order used by the template (or the setting of the |af= parameter once proposed by Headbomb to override the default order); at present (and without |af=), this would be "<last>, <first>" (because the format used in the |-mask= parameters is most likely needed to be in the same order as in the normal display of author names without |author-mask=). A doubled ** would result in the opposite order of the one selected by *; at present, this would be "<first> <last>". Triple *** would result in only the <last> name. Again, if someone uses *, ** or *** in |author-maskn= in conjunction with |authorn= rather than |author-firstn=/|author-lastn=, the value of |authorn= is taken as a replacement.

So much for the overview. There are a number of special cases not discussed yet, but I think you already get the idea: One easy to remember global placeholder * as a generic smart and context-specific placeholder, possibly doubled ** or tripled *** to switch the output format depending on context (similar to different signatures being issued with ~ ~~ ~~~ ~~~~ or different link types being used depending on if using single or doubled brackets etc.) Since, in the context of each parameter, it is quite obvious what would be a reasonable source for the substitution (AFAI see it, there is always only one source which really makes sense), this scheme is easy to remember and use. It would be backward compatible and optional to use, it would reduce the amount of required manual input, reduce the risk for many typos, may make many citation templates easier to read on source code level (YMMV) without having to reduce the amount of provided info, and it would even save some storage space.

--Matthiaspaul (talk) 14:48, 8 August 2020 (UTC)

New autolinking of free dois breaks citations that deliberately omit title

In the citation

{{cite journal|last=Nespolo|first=Massimo|date=November 2019|doi=10.1107/s1600576719014055|issue=6|journal=Journal of Applied Crystallography|pages=1467–1468|title=none|volume=52|doi-access=free}}

currently formatted as

Nespolo, Massimo (November 2019). Journal of Applied Crystallography. 52 (6): 1467–1468. doi:10.1107/s1600576719014055.{{cite journal}}: CS1 maint: untitled periodical (link)

I am now seeing a visible url and a big red error "|doi= missing title" even in the not-logged-in view. This citation was previously error-free and was broken by recent changes to the citation template that added auto-linking of dois without taking adequate precautions that there is something to auto-link to.

I frequently and deliberately omit titles, using the documented method |title=none of doing this, as part of lists of multiple reviews of the same book that do not have individual titles. Many such reviews are not formatted with titles so any title that one included would have to be a made-up placeholder. In this case the review actually is formatted with a title, but not one that would be useful to include in a citation: the title, as formatted in the review, is "Fat Chance! Probability from 0 to 1. By Benedict Gross, Joe Harris, and Emily Riehl. Cambridge University Press, 2019. Pp. Xi+200. Paperback price GBP 19.99, ISBN 9781108728188. Hardback price GBP 49.99, ISBN 9781108482967. Ebook price USD 21.00, ISBN 9781108598705."

It should not be an error to omit the title and it should not be an error to mark dois as free even when the title is deliberately omitted. Please stop making the citation templates even more brittle and unusable than they have become, restore the ability to deliberately omit titles on citations, and un-break the many citations already existing within Wikipedia articles that this change has broken. —David Eppstein (talk) 22:18, 4 August 2020 (UTC)

This seems like a fairly clear bug to me. The tone from "please stop" was not necessary accordingly. --Izno (talk) 22:49, 4 August 2020 (UTC)

{{cite journal/new|last=Nespolo|first=Massimo|date=November 2019|doi=10.1107/s1600576719014055|issue=6|journal=Journal of Applied Crystallography|pages=1467–1468|title=none|volume=52|doi-access=free}}

Nespolo, Massimo (November 2019). Journal of Applied Crystallography. 52 (6): 1467–1468. doi:10.1107/s1600576719014055.{{cite journal}}: CS1 maint: untitled periodical (link)

—Trappist the monk (talk) 22:55, 4 August 2020 (UTC)

Thanks! It has historically not always been easy to distinguish changes that are recognized as bugs and fixed from changes that are intended to restrict how the template can be applied and are permanent. I'm glad this one landed on the fixed side. —David Eppstein (talk) 23:11, 4 August 2020 (UTC)

Perhaps some common sense would be useful next time? − Pintoch (talk) 07:08, 5 August 2020 (UTC)

Us editors on this talk page have been accused of lacking such on occasion. --Izno (talk) 18:54, 5 August 2020 (UTC)

Common sense is not as common as people think. It may also be not as sensible. 65.88.88.69 (talk) 17:26, 7 August 2020 (UTC)

Towards solving pending issues of the auto-link feature...

This was one of the (thankfully now fixed) open special cases already mentioned in the original thread (Help talk:Citation Style 1#Proceeding). There are more (see also Help talk:Citation Style 1#module suite update 11–12 July 2020). Would be nice to see them addressed before the next update is about to be rolled out, at least the manual override/disable facility we agreed upon by overloading |url=none/doi/pmc/<other-identifier-parameter-name>/<url>... --Matthiaspaul (talk) 12:26, 8 August 2020 (UTC)

If it was not clear, I oppose the use of |url=none/doi/pmc/.... Introducing these non-URL values in a field which currently expects URLs is likely to break many tools and scripts - this is a bad programming practice. Beyond that, I am yet to see an example of a citation where a |url=none would be desirable. For |url=doi and others, I would simply let people add those URLs directly in the URL field - there is no need for the additional complexity and indirections. If bots currently remove such URLs, they should stop doing that. − Pintoch (talk) 17:31, 8 August 2020 (UTC)

I agree that populating the url field with non-url content is not a good practice. I would make an exception for |url=none as long as the current dependency of |archive-url= is in place. I am fairly certain that reliable archives may have fascimiles of originals that can no longer be found. 98.0.246.242 (talk) 00:04, 9 August 2020 (UTC)

Alright then, in the previous discussions I raised specifically this concern as a potential problem twice, but over the course of weeks nobody confirmed it as a real problem, so weighing the various pros and cons we more or less settled on |url=none/doi/pmc/... (and reserved |chapter-url=none/doi/pmc/... for potential future use).

Avoiding this and leaving |url= (and |chapter-url=) for actual URLs only, the alternative is to overload |title-link=none/doi/pmc/..., instead. (In this case we'd have to support our special "accept this as is" ((syntax)) for |title-link= to still allow linking to Wikipedia articles in the rare case where they might clash with the bunch of supported identifier parameter names (no problem for the current ones, but could be if the list gets longer, as already requested) - alternatively, in these few cases, we could go through a redirect to avoid the name conflict. So, not a real issue, either.) However, if, at some point in the future, we'd add auto-linking also for chapters (as was already requested as well), the user interface might no longer remain self-expaining, as there is no corresponding |chapter-link= parameter (and no need to add one, as we don't have Wikipedia articles on chapters, or do we?). So, if we would want to override the automatic behaviour in this case, we would have to add these cases as arguments to |title-link= in a sensible way as well - it's definitely doable, the only case we could not control then is when a title is linked to a Wikipedia article (thereby occupying the parameter) and the chapter title should be linked to something different than the automatic default (or not at all).

The third alternative is to have a separate parameter to control the auto-linking behaviour like |auto-link=auto/none/doi/pmc/.... However, I don't like this variant much because it is bad user-interface design (and therefore difficult to remember for users) to scatter related functionality over multiple parameters in particular if they cover mutually exclusive cases like |auto-link=doi |title-link=WP-Article which requires error checking for cases (like "Conflicting link targets defined.") which could be ruled out simply by the underlying parameter logic already (because one parameter can hold only one value). The idea is to make existing parameters smarter rather than to introduce yet more new parameters for special cases which could be covered as part of existing parameters as well.

--Matthiaspaul (talk)

In my opinion, all these proposals try to address a problem that does not exist. Editors can link citation titles with |url= or |title-link= to any URL or internal link, there is no need to introduce a complex machinery to provide a convoluted alternative. The simpler we can keep these templates, the more reliable and easier to maintain they will be. I still have not seen an example of a citation where disabling auto-linking is desirable, too. − Pintoch (talk) 15:07, 9 August 2020 (UTC)

trans-quote?

Hi, I feel like I remember a parameter trans-quote for translated quotes from citations in foreign languages. Am I remembering incorrectly, or has this been removed? Thanks, Ezhao02 (talk) 13:31, 5 August 2020 (UTC)

Never existed. |quote= is a free-form parameter so you can include translations in it if you would like. Better in my mind, if the quoted material is important to the article, put that material in the article, translate it there and cite both; don't clutter the references section with quotations.

—Trappist the monk (talk) 13:56, 5 August 2020 (UTC)

Thanks for the information and the advice. Something must have happened to my memory. Ezhao02 (talk) 14:01, 5 August 2020 (UTC)

You probably have seen one of the proposals to add this (and |script-quote=) in the past. While Trappist is right that |quote= is a free-form parameter so you can add your own formatting, it would be desirable to have a consistent style centrally maintained instead of every editor having to invent his own conventions for this. Therefore, such a parameter is desirable. --Matthiaspaul (talk) 12:03, 8 August 2020 (UTC)

Wayback Machine

Hello, I came across a reference title of "Wayback Machine", having done a search there are close on 2,000 of these. Was wondering if it worth extending the "Archived copy" tracking to include this as well, so that the articles can be modified to give something useful. Keith D (talk) 21:21, 5 August 2020 (UTC)

This would be good. There is a tool (reFill) that when it encounters a bare or square web.archive.org link generates a {{cite web}} set with |title=Wayback Machine and |website=web.archive.org (diff). My tool WaybackMedic unwinds some of it (diff) but the title remains as "Wayback" -- GreenC 01:47, 6 August 2020 (UTC)

Why does Wikiquote's Ref template and Wikipedia's Ref template handle commas in the date field differently?

(X-posting from the Tech Pump) Wikiquote's ref template will parse "July 2 1999" just fine, but our template requires a comma, e.g. "July 2, 1999". Why is that? Can someone fix our template to stop caring so much? I screw this up on my first pass something like 25% of the time. -- Kendrick7^talk 00:30, 7 August 2020 (UTC)

"July 2 1999" is not a valid date format here on the English Wikipedia, per MOS:DATESNO, so our CS1 citation templates (e.g. {{cite web}} and similar) display an error message. Wikiquote does not have enforceable date styles (per the English Wikiquote MOS), so ambiguous and non-standard dates aren't marked as errors. – Jonesey95 (talk) 01:58, 7 August 2020 (UTC)

What does MOS have to do with it? Wikiquote's template simply inserts the comma automatically, so the styling result is exactly the same. -- Kendrick7^talk 12:37, 7 August 2020 (UTC)

Which template is doing this formatting? See my wikiquote sandbox, where Template:cite news on wikiquote is leaving "July 2 1999" alone. You are welcome to add examples to that sandbox page. – Jonesey95 (talk) 15:43, 7 August 2020 (UTC)

Wikiquote's version of "cite web" is definitely adding the comma; I've added an example to your sandbox. I don't know why I expected consistency, lol. -- Kendrick7^talk 00:22, 8 August 2020 (UTC)

Help

I have a question about a book that is divided into several individual sections with their own authors. ¿How should I cite that book? In my case I am only interested in one section of that book. --Muwatallis II (talk) 23:11, 8 August 2020 (UTC)

Author. "Section". Title. {{cite book}}: |author= has generic name (help) is the basic skeleton. If you have some more information we can fill out the rest for you. --Izno (talk) 23:12, 8 August 2020 (UTC)

The book is called: Escuadra Nacional 1818-2018 (no author, only publisher)

The title of the chapter or section of the book already mentioned is: De la Guerra del Pacífico hasta fines del siglo XIX. The author is: Piero Castagneto Garviso. This author is just from the chapter or section of the book.

As I mentioned before, the book has other chapters or sections with their own authors, but I'm only interested in the one already mentioned. --Muwatallis II (talk) 00:04, 9 August 2020 (UTC)

Piero Castagneto Garviso. "De la Guerra del Pacífico hasta fines del siglo XIX". Escuadra Nacional 1818-2018. Publisher. is how that is done indeed. I expect there is an editor too who you might want to include, and of course a year. --Izno (talk) 00:25, 9 August 2020 (UTC)

+|url=http://anyflip.com/yccc/bhap:

Piero Castagneto Garviso. "De la Guerra del Pacífico hasta fines del siglo XIX". Escuadra Nacional 1818-2018. Publisher.

—Trappist the monk (talk) 00:30, 9 August 2020 (UTC)

Placement of language annotation

When the source being cited is a chapter in an edited book or an article in a journal, the language of the source (specified with |language=) should be notated after the chapter or article, rather than after the book or journal, which may include items in several languages. For example, currently we have:

• Author (2020). Buch (in German). {{cite book}}: |author= has generic name (help)
• Autor (2020). "Kapitel". In Editor (ed.). Book (in German). {{cite book}}: |editor= has generic name (help)
• Autorin (2020). "Artikel". Journal (in German). 3: 1–20.

The first is fine, but in the last two "(in German)" should be moved forward, after the chapter or article. Kanguole 11:50, 9 August 2020 (UTC)

Recently discussed here; see particularly my comments. --Izno (talk) 11:59, 9 August 2020 (UTC)

It seems there was substantial support for putting the annotation after the title of the item, with a few in favour of putting it at the end. Either would be better that the current practice of placing it after the name of the book, journal or series, which may be in several languages. But then the discussion got derailed onto formatting of multiple parentheticals. Kanguole 14:12, 9 August 2020 (UTC)

The conclusion you should draw is that the problem is not easy because of the paranetheticals. --Izno (talk) 14:47, 9 August 2020 (UTC)

Request for new maintenance category for abbreviated year ranges in the date= parameter

Hi. Non-abbreviated year ranges are our preferred format for year ranges and there is certainly no particular need to support abbreviated year ranges in citations (except for if we can) - they could certainly be written in non-abbreviated form as well. Year ranges are comparably rare in citations, even more so abbreviated ones, as in most cases the publication date specifies a specific point in time rather than a span. I have seen less than a handful of abbreviated year ranges in citations in all those years.

On the other hand, incomplete dates consisting of only the month and the year and no day are very common in citations (I see them every day), but in the case of the ymd date format, the form "yyyy-mm" is disallowed in order to avoid a possible confusion with "yyyy–yy". Since the EDTF form "yyyy-mm-XX" is not currently supported as well (would be useful at least on source code level, not for display), this leads to such dates being rewritten as "Month yyyy", which unnecessarily creates inconsistency when all the other dates in the citations are given in ymd format.

Hence, it is reasonable to swap this around and fade out and ultimately disallow abbreviated year ranges in the |date= parameter of citations at some point in the future (only there, not elsewhere where they are still allowed, including f.e. in |title=, |chapter=, or |quote= parameters), so that, at some further point in the future, we can officially allow "yyyy-mm" in citations already using the ymd format. (In order to keep the change as minimal as possible, this is meant to affect only citations, not dates in tables or in the article body.)

In order to get a grip on how many citations actually have dates formatted this way at all, I would appreciate a tracking/maintenance category for citations using any detectable form of abbreviated year range (but at least the "yyyy–yy" form). --Matthiaspaul (talk) 13:19, 9 August 2020 (UTC)

Can you please give a list of examples of what you consider abbreviated year ranges? It is unclear what you consider acceptable and unacceptable.

Our date format checking follows MOS:DATESNO and MOS:DATERANGE; is there an invalid format listed there that is not being detected? – Jonesey95 (talk) 15:01, 9 August 2020 (UTC)

Here's an example of an invalid range that is currently detected as an error: Title, 1902–07 {{citation}}: Check date values in: |date= (help). – Jonesey95 (talk) 15:03, 9 August 2020 (UTC)