This is related to T413679. It is currently (slightly) ameliorated by T382446 (now deployed).

This is indeed how it used to work; the behaviour was changed upon user request. T124610

It works best if there is a url or doi in the reference, but it can also work with ones that don't as long as it's a scholarly reference, via crossref. It also worked better when we had worldcat because then we could do book refs.

Rather than removing it completely, I would explore first

• Changing the UI so the user can see the diff and check that the new citation matches the old one (potentially slow if this is urgent)
• Adding back in better search functionality. This could be done with worldcat (T352571) or open library. worldcat (T413785) is more comprehensive, but open library is free and also more expedient. (This doesn't really help in the screenshotted cases, but would in other cases.)

@Mvolz to be doubly-sure I'm following, could you please...?
A) Let me know what (if anything) I've missed/misunderstood about what you shared in T413279#11503238?
B) Say a bit more about how (if at all) you see, "adding back in better search functionality" relating to the "Convert reference" suggestion?

What I'm understanding

1. This ticket is related to T413679 in so far as the latter would, theoretically, enable people who act on this "Convert reference" suggestion the feedback necessary to verify whether the reference they converted was done accurately and without losing data?
2. Limiting the "Convert reference" functionality to just plain URLS to start could be viewed as a step "back" in so far as this behavior was intentionally changed via T124610
3. Rather than limiting the scope of the "Convert reference" suggestion to just plain links to start, what if we could A) add a step within the worfklow that would enable people to confirm, as you put it, "...the new citation matches the old one." or B)

Change #1226397 had a related patch set uploaded (by DLynch; author: DLynch):

[mediawiki/extensions/VisualEditor@master] ConvertReferenceEditCheck: require the link annotation to cover the ref

https://gerrit.wikimedia.org/r/1226397

To elaborate on what that patch does...

Given this wikitext:

REFERENCE ONE: <ref>Author One, Author Two. [http://google.com/ Book Name]. p.58.</ref>.
REFERENCE TWO: <ref>[http://google.com/ Book Name]</ref>
REFERENCE THREE: <ref>https://google.com</ref>

Before the patch, all of these would get offered the conversion. After the patch, only TWO and THREE would be offered it.

This could be expanded further to require that there not be any link-text, just the URL. In that case, only THREE would be offered a conversion.

In T413279#11519511, @ppelberg wrote:

@Mvolz to be doubly-sure I'm following, could you please...?
A) Let me know what (if anything) I've missed/misunderstood about what you shared in T413279#11503238?
B) Say a bit more about how (if at all) you see, "adding back in better search functionality" relating to the "Convert reference" suggestion?

What I'm understanding

1. This ticket is related to T413679 in so far as the latter would, theoretically, enable people who act on this "Convert reference" suggestion the feedback necessary to verify whether the reference they converted was done accurately and without losing data?
2. Limiting the "Convert reference" functionality to just plain URLS to start could be viewed as a step "back" in so far as this behavior was intentionally changed via T124610
3. Rather than limiting the scope of the "Convert reference" suggestion to just plain links to start, what if we could A) add a step within the worfklow that would enable people to confirm, as you put it, "...the new citation matches the old one." or B)

Sorry about that, this is me not understanding the context of the ticket (and getting messaged on google chat about meetings I'm not actually in! haha)

I think limiting the scope in edit check is great, I had misinterpreted this ticket as being about the general experience.

In T413279#11522080, @Mvolz wrote:

Sorry about that, this is me not understanding the context of the ticket (and getting messaged on google chat about meetings I'm not actually in! haha)

I think limiting the scope in edit check is great, I had misinterpreted this ticket as being about the general experience.

All good and great to know we're aligned. Thank you for clarifying, @Mvolz.

Meta: I too have noticed myself confused at times by the Google Chat notifs.

Per today's offline team meeting, we're limiting the scope of this initial suggestion to plain links, for now.

Where "plain links" here means there is not any link-text, just the URL. E.g. <ref>https://google.com</ref> as @DLynch described in T413279#11519612.

Thinking: we assume the approach described above poses a lower risk of people inadvertently removing useful metadata by way of the suggestion. We recognize in taking this approach, we are increasing the number of false negatives, and by extension, lowering the likelihood that people will encounter this suggestion. If/when we come to find the limited prevalence of this suggestion being an issue, we'll consider expanding the scope of the suggestion via a ticket like T414027.

Change #1226397 merged by jenkins-bot:

[mediawiki/extensions/VisualEditor@master] ConvertReferenceEditCheck: introduce strictness configuration

https://gerrit.wikimedia.org/r/1226397