Wikipedia talk:Wikidata/2018 State of affairs

Archives

Archive 1

This page has archives. Sections older than 14 days may be automatically archived by Lowercase sigmabot III when more than 3 sections are present.

Preceding discussions at Wikipedia talk:Wikidata/2017 State of affairs (and archives)

The RfC on the short descriptions magic word (archived at Wikipedia:Village_pump_(proposals)/Archive_145) has been closed. The result is

"The consensus is #5 for the first question - To populate the magic words by starting with blanks, and allowing them to be filled in manually and/or by bot (as per usual bot procedures). The consensus is #2 for the second question - Show no description where the magic word does not exist. "

Can the WMF please indicate whether they will respect the result of this RfC and when we may expect the magic word to be available? Fram (talk) 10:44, 7 February 2018 (UTC)[reply]

• I guess now we'll see whether the WMF will shut off almost all short descriptions for mobile users probably indefinitely, on a mere 10-6 vote. I rather hope they don't. @DannyH (WMF): ? Jheald (talk) 11:07, 7 February 2018 (UTC)[reply]
  • You (one of the "6" of that vote, let's not restart that RfC here) are free to populate the magic word once it is available, or to request a bot to populate it. The rfC was not "we will not use the magic word" or "no short descriptions". And of course, it wasn't the first RfC on this, we already had a previous RfC which supported the shutdown of the Wikidata descriptions on mobile, this is just a confirmation for the WMF that yes, this is what we really want. Fram (talk) 11:12, 7 February 2018 (UTC)[reply]
    • Meh. 4 !votes isn't much of a mandate. I think WMF would be entirely within its remit as steward for the interests of our users to say: fine, we note this vote, we will turn off descriptions drawn from Wikidata -- once the community has demonstrated it has the capability and will to step in, and that it will not leave mobile users hanging in the lurch, eg once it has created local descriptions for at least 50% of articles. That would seem an entirely reasonable and discriminating stance for the WMF to take.

    As for myself, thank you very much, but I already have a full slate of sources I'm working on, sticking to adding new stuff that's actually useful, rather than wasting my time pointlessly duplicating content we already have. But one thing I would say, if people are thinking of automated creation of local descriptions here, is that it probably makes sense to go via improving the descriptions and/or auto-descriptions that are on Wikidata first. The nature of Wikidata makes these easier to extract for particular large subsets of articles, to understand their patterns, adequacies, and deficiencies in bulk, and easier to then iteratively re-extract and re-update, to evaluate at scale against conformance for particular goals. If the aim is collaborative assessment and improvement, it's a platform that provides the capabilities needed for the work. Certainly better to put in what fixes can be made there first, rather than to blindly import whatever WD may have at the moment to here as an initial step, where it will probably never be systematically looked at, and never fixed. Jheald (talk) 12:21, 7 February 2018 (UTC)[reply]

    • • They have never (or cerainly not adequately) checked whether the Wikidata community (who currently "maintain" these descriptions" had the "capability and will to step in", so imposing this as a conditio sine qua non on enwiki seems rather one-sided. The WMF is rarely if ever a "steward for the interests of our users", usually they are a steward to the interests of their paid developers and myriad of managers and intermediaries only. Episodes where the WMF imposed its will against the wishes of the community have invariably ended in strife and have created lots of distrust and disdain, and in the end usually resulted in the WMF having to back down anyway (Gather, Flow, ...). And you still don't seem to get that it is not within our remit to go and "fix" the Wikidata descriptions on Wikidata, as many of them are perfectly suited for their real purpose (a description on Wikidata) but unfit to be used on enwiki. To change these to descriptions suited for enwiki would not be welcomed (and rightly so). We are two different projects, and using one to populate the other is in most cases a bad idea (e.g. the millions of "imported from Wikipedia" sourced items on Wikidata, which are now being imported back into Wikipedia as if they are reliable information). Fram (talk) 12:44, 7 February 2018 (UTC)[reply]
        • I'm not saying you have to do it this way. I'm just saying, if you want to populate local descriptions on en-wiki, going this way and using Wikidata as a staging area probably makes sense.

        First, though, you need to create a style guide for the descriptions you want on en-wiki -- ie how long they should ideally be, what information should be included if available for various different types of article, etc.

        Then, as I presume you won't be literally creating 5.5 million descriptions from scratch, you need to come up with sources of candidate descriptions -- perhaps what's already at Wikidata, or one of Magnus's auto-descriptions based on Wikidata, or some other source; or perhaps multiple such sources, making a bespoke choice or combination per article.

        Then you need to evaluate whether those descriptions actually match your style guide -- do they contain the key information they should do?

        Maybe not all of it, and maybe not for every article, but a lot of this work it would make sense to do in a Wikidata environment -- because Wikidata gives you a space where relevant information is available, and where it is easy to query and retrieve the texts and properties for large related groups of articles at scale.

        Yes, there will be some cases where en-wiki's needs aren't precisely the same as Wikidata's (eg disambiguation pages?); but mostly I imagine descriptions in line with whatever style guide en-wiki will come up with would be entirely acceptable on Wikidata (wd's requirements are probably not as exacting in this area as en-wiki's) and I would think that descriptions that had been systematically analysed and raised up to the en-wiki style guide for information inclusion and conciseness would be very welcome.

        You may see all the Wiki projects as entirely separate. I don't. I see the relationships as much more synergistic. If I create a tracking table like Talk:List of Royal Academicians/RAs it's because it makes sense to make a hub for improving all of the projects together -- which RAs don't have Wikipedia articles? Which don't have Commons categories? Which might we identify with more external IDs? Sometimes different projects need to go their own way, and that flexibility is valuable. But it makes little sense to create walls where they don't need to exist. Jheald (talk) 13:29, 7 February 2018 (UTC)[reply]

        • (ec)I have no objection to Wikidata eventually using our descriptions if they are so inclined, that's their choice (and usually they will be so short and simplistic that copyvio / need for attribution seem to me like something of no concern in this case, although others have raised this as an issue when one would copy in either direction). As for your "improving all of the projects together" and "which RA don't have Wikipedia articles", I think you refer to enwiki alone, and not the 200+ other languages? In any case, no one is stopping you from doing this (I hope), but I have no interest in dividing my limited time over multiple projects, with their own policies, rules, goals, ..., and I see it as seriously detrimental to divide the actual textual article content over multiple, poorly integrated sites, as it seriously complicates editing, maintenance, ... for little actual benefit (Wikidata is not in general more up-to-date or reliable than what we have here, and has much looses sourcing rules). Using Wikidata to generate project lists, talkpage lists, ... all kind of checks and balances, fine! That's a thing it can be really useful for, e.g. lists of missing articles and so on. But not for autogenerating actual articles and mainspace lists, or populating infoboxes, or showing texts to readers which isn't maintained at enwiki. We shouldn't throw away the strength, the core of Wikipedia just because we have a shiny newer tool. Fram (talk) 13:50, 7 February 2018 (UTC)[reply]

          • Fram, I am just saying that practically, as a matter of workflow, if you're serious about creating these vast numbers of description texts, with any form of group quality control, it probably makes sense to use Wikidata as a staging area, because of the technical capabilities that would give you. Jheald (talk) 14:04, 7 February 2018 (UTC)[reply]

            • I don't edit Wikidata and have no intentions of doing so. Too many toxic personalities who have been shown the door or are seriously restricted here are given free reign there. Furthermore, the policies (or near-total lack of them) and anything-goes mentality (be it unsourced articles about non-notable minors, or the use of sources which are completely unreliable and full of copyvios) turn it into a place I don't want to be a part of. I have no idea how you would actually use Wikidata as a staging area unless you propose overwriting their descriptions with whatever we want. Like I said, you (or anyone) is free to make such suggestions at the bot owners noticeboard or at some village pump. I don't see how this would be practical or beneficial, seems much easier to e. copy all Wikidata descriptions to the corresponding article talk pages and get some approval process there, but it isn't up to me to decide if, when and how we will populate the magic word. Fram (talk) 14:11, 7 February 2018 (UTC)[reply]

              • My point is, if as an initial step you're going to copy all Wikidata descriptions to the corresponding article talk pages, then before you do that, as a first stage, it makes sense to fix deficiencies and raise those descriptions as far as you can closer to what you are looking for in a systematic way on Wikidata first, because to make those changes on a mass scale is far easier at that stage than later; and to do it in a way that is open to collaboration. If going to do that first draft as a machine process, then do it on Wikidata, because the very design of Wikidata is created to be easier for machine processes like that -- easier to read on a mass scale, easier to write on a mass scale, easier to analyse on a mass scale. Jheald (talk) 14:52, 7 February 2018 (UTC)[reply]

        Wiki projects have different communities and different standards. It would be inexcusably offensive to overwhelm the community at Wikidata to enforce English Wikipedias norms, and it would be equally inexcusable to allow content to be imported against the requirements of English Wikipedia's policies, particularly regarding BLPs, medical information and other things which we hold ourselves accountable for and which may have real life consequences. Using Wikidata just does not cut it. Wikipedians are accustomed to the Wikipedia editing environment and can usefully watchlist their areas of interest. Wikidata is a foreign territory, with foreign customs to most Wikipedians. This is not a fault of Wikidata, it is because they are not Wikipedia. Some of us are both Wikipedians and Wikidatans, but not enough for it to work. Short descriptions are text, not data - they are different in different languges, and the articles they describe are different too. In some cases the articles have different scope, so there is doubt whether the same data item realistically applies. If English Wikipedia has an article that uses the same Q number as another article in another Wikipedia, and the subject is different, which one gets priority? Even Wikidata does not have a procedure for settling such a problem. There is no definitive scope for a Wikidata item as far as I can find out. On Wikipedia we can adjust the short description to match the article as and when the scope changes. Changing it on Wikidata takes no account of how it relates to other articles in other languages, which may change in different directions. · · · Peter (Southwood) ^(talk): 14:13, 7 February 2018 (UTC)[reply]

        I'm not talking about using Wikidata as the final resting places for these descriptions, I'm talking about using it as a staging area where draft descriptions can be created and reviewed on a mass scale -- because you are going to need to create millions of these things. It's going to need to be done and quality-controlled systematically, as a driven process. Just leaving it to haphazard meanderings on individual talk pages is not going to cut it -- if you go that route you still wouldn't have even a fraction of the 5.5 million you're going to need, even after a decade.

        It's all very well for people to talk airily here about what may or may not be acceptable. But will the capability and commitment of the community exist to actually implement an alternative? Talk is all very well, but who will commit all the hard graft to make it actually happen? Will the 10 who voted for it be able to achieve what they have voted for? WMF would be very sensible to see some evidence on that score, before dismantling the existing capability for all its imperfections. Jheald (talk) 14:39, 7 February 2018 (UTC)[reply]

    [ec]Consensus closure is not suppoed to be a count of votes, !votes or any other form of numerical majority. It is supposed to be about the quality of arguments made in support or opposition as well as an indication of general opinion. The closure is not quite my preference, but it does not appear to be wrong. If you think that the closure does not follow the consensus, take it up at the RFC. Otherwise we are held to this decision and we expect WMF to go with it as well.

    I too would welcome an update on the progress of the magic word. At the least I would like to see the syntax defined, so we can start entering short descriptions with the correct format. That would give the WMF the desired information sooner rather than later, and as I understand it, the RFC requires that WMF should stop using Wikidata (short descriptions^{(correction added)}) to identify English Wikipedia articles in any way, anywhere. · · · Peter (Southwood) ^(talk): 13:47, 7 February 2018 (UTC)[reply]

    @Pbsouthwood: "the RFC requires that WMF should stop using Wikidata to identify English Wikipedia articles in any way, anywhere" - no, that's not the question that was asked. There is no such requirement from an RfC. Thanks. Mike Peel (talk) 14:13, 7 February 2018 (UTC)[reply]

    Mike Peel, you are right, I left out short descriptions. · · · Peter (Southwood) ^(talk): 14:31, 7 February 2018 (UTC)[reply]

    Actually we already had an RFC that covered that. Its only because the WMF decided to sneakily not do what was asked that we have ended up at the magic word compromise. What there has been consistently consensus for is for Wikipedia articles to not have associated descriptions that are based elsewhere, edited elsewhere and out of the control of Wikipedia editors, and displayed as if they are part of the Wikipedia article. Either the descriptions for ENWP articles are under the control of ENWP, or they are not displayed. Only in death does duty end (talk) 14:15, 7 February 2018 (UTC)[reply]

    @Only in death: As in this RfC? That was withdrawn, it was never completed. I'm just pointing out that there is not an RfC decision on this, that's all. Thanks. Mike Peel (talk) 14:21, 7 February 2018 (UTC)[reply]

    • It was closed because the WMF promised to stop using Wikidata, it wasn't closed as no consensus. Using the lack of formal closure, caused by the WMF promise, as a reason to claim that that result wasn't decided, seems, well, let's settle for "unfair". And in any case, the RfC now concluded with Show no description where the magic word does not exist which seems to be what Only in death said, "stop using Wikidata to identify English Wikipedia articles". No magic word or blank magic word = no description, magic word populated on enwiki = show enwiki description. There is no room in that conclusion to add a "show Wikidata description" somehow. Fram (talk) 14:26, 7 February 2018 (UTC)[reply]

      • That's a rather creative way of interpreting RfC results (or non-results as the case might be). Mike Peel (talk) 14:30, 7 February 2018 (UTC)[reply]

        • ? Not much creativity involved, actually. If people want to continue showing the current Wikidata descriptions, they will have to add these to our articles (by bot presumably). But the closure of the RfC indicates that we will no longer display the Wikidata descriptions directly on enwiki (mobile, search, whatever). That seems rather clear from the RfC closure. Fram (talk) 15:03, 7 February 2018 (UTC)[reply]

          • That is not at all my understanding of the outcome. Once the magic word is available and in action, then yes. But right now, we haven't got the magic word - and you need to generate a few million descriptions for it before it is turned on the second part of the RfC comes into action... Mike Peel (talk) 16:27, 7 February 2018 (UTC)[reply]

            "you need to generate a few million descriptions for it before it is turned on" Iff that is the case, I predict a riot. No one is going to populate a few million articles with a magic word when this has no effect at all until the WMF may decide that our effort is sufficient, or not. I don't know where you get these ideas, and whether you believe them to be good ideas or not, but this is possibly the worst way to deal with this at all. The WMF proposal was to only switch off the Wikidata descriptions as default / backup for missing or empty magic words after some time and/or number of descriptions had been generated. You propose to provide us with a magic word which won't do anything until millions of them have been added and filled. Which is not what the WMF wants (or at least, not what they say), and not what the RfC outcome wants. And is not helping the discussion at all of course. Fram (talk) 16:50, 7 February 2018 (UTC)[reply]

            • Sorry, I mis-spoke, I've struck the last bit and clarified it. As I've said repeatedly, I think a magic word is a daft idea in the first place, so none of this is a good idea. Mike Peel (talk) 17:10, 7 February 2018 (UTC)[reply]

              Hi all. WMF's intention is to do what we talked about in late December/early January. We're building a magic word -- the Phabricator ticket is phab:T184000, and the developers have started work on it. There are two stages. In the first stage, descriptions posted on English Wikipedia using the magic word will be displayed as the short description for that article. On a given article, if there isn't a magic word or the description is blank (broadly construed, including blank spaces) -- then it falls back to the Wikidata description as a default. The second stage starts when the English WP editors have filled in enough descriptions that it's roughly comparable to the existing Wikidata descriptions (around 2 million descriptions). At that point, we switch to only using the Wikipedia magic-word descriptions, and won't use the Wikidata descriptions as a backup. -- DannyH (WMF) (talk) 19:12, 7 February 2018 (UTC)[reply]

              @DannyH (WMF):--Hmm...In that case, you are not respecting the result of the 2nd part of the RFC or am I missing something? Regards ~ Winged Blades^Godric 03:46, 8 February 2018 (UTC)[reply]

              Oh! I missed your earlier statement--.....we're not going to agree to a course of action that results in the mass blanking of existing descriptions, for any meaningful length of time. Best, ~ Winged Blades^Godric 03:49, 8 February 2018 (UTC)[reply]

              Who is this "we" of whom you speak?· · · Peter (Southwood) ^(talk): 14:07, 8 February 2018 (UTC)[reply]

              @Pbsouthwood:--Obviously, Danny.~ Winged Blades^Godric 09:22, 9 February 2018 (UTC)[reply]

              • All one of them? · · · Peter (Southwood) ^(talk): 13:21, 10 February 2018 (UTC)[reply]

Yes, Danny, we know your intention. Which isn't what the enwiki community wants. Wy is it so unacceptable to get rid of the descriptions if enwiki doesn't want them as is (and will populate them if and when they want to), but apparently it is no problem that mobile viewers lose data from articles (e.g. navboxes and categories)? It seems highly hyprocritical and fundamentally wrong from the WMF to impose their content decisions (which parts can be hidden from view and which parts need to be there at all costs) when that is not the role of the WMF. The WMF should make things available which are up to the communities to use or not, and the WMF should do their best to make sure that content is visible for most users on most devices. If some content is not provided by a project (like enwiki), then the WMF should not provide it from elsewhere, nor should they force the project to provide such content. Fram (talk) 10:13, 8 February 2018 (UTC)[reply]

This brings me back to a question I have asked before, but has not yet been answered. I will try to make it simple and unambiguous, and request a simple and unambiguous answer. DannyH (WMF), Who is responsible for the decision to continue displaying Wikidata short descriptions for English Wikipedia articles on mobile and other interfaces? If it was one person, please provide their name, If it was a committee decision, please name the committee and chairperson. · · · Peter (Southwood) ^(talk): 11:26, 8 February 2018 (UTC)[reply]

Pbsouthwood: There were a lot of people from different teams involved in the decision. The most visible use of the short descriptions is on the iOS and Android apps, so that team was a big part of it, along with other people from the Readers team and Community Engagement. I took the lead on communication and figuring out the compromise solution, so I am the person most responsible for that decision.

Responding to Peter and Fram's questions: The short descriptions are useful for the readers and editors that use them, and we're not willing to mass blank them, even with the RfC consensus. It would be bad for those users, and those products. But I agree that Wikipedia editors should have editorial control over the descriptions, and that's why we're building the magic word, and why we'll switch to entirely Wikipedia-generated descriptions once there are enough that the current users won't notice the change. Sorry for just repeating things I've said before, but it's the same question, so I still have the same answers. -- DannyH (WMF) (talk) 19:35, 8 February 2018 (UTC)[reply]

@DannyH (WMF): Categories and navboxes and so on are useful for the readers as well, but the WMF had no problem with "mass blank" them on mobile view either. Talk pages were nearly inaccessible from mobile view for years as well. It seems like you use that argument only when it suits you, but ignore it otherwise (just like you ignored it here). Fram (talk) 19:41, 8 February 2018 (UTC)[reply]

Fram: That's actually been the subject of some conversations at the WMF as we're working on the annual plan for next year (July 2018-June 2019). We haven't finalized the plans yet, but the mobile and apps teams are talking about helping people edit on both the mobile web and the apps. That means providing more functionality on those platforms, including better notifications and access to talk pages. I'm not sure about categories or navboxes, but I can ask. We'll have a better idea of what the plans are in a couple weeks, and then various teams will be having conversations on wiki about what people need, and how to provide it. -- DannyH (WMF) (talk) 19:54, 8 February 2018 (UTC)[reply]

But meanwhile it wasn't considered a problem to remove this functionality from the readers in mobile view (while adding the rather rubbish "related articles" instead). I'm just trying to understand why if it suits the WMF, it is acceptable to remove content, functionality, helpful links from articles, but when it suits enwiki, it suddenly becomes unacceptable. Fram (talk) 19:58, 8 February 2018 (UTC)[reply]

Possibly because WMF makes the rules for what they do to suit their goals without consulting the projects, or by interpreting feedback from the projects to suit their own agendas. Or possibly because the WMF departments interpret what policy may exist to suit themselves. Or some other reason no-one could be bothered to explain. Who knows... · · · Peter (Southwood) ^(talk): 08:10, 16 February 2018 (UTC)[reply]

Returning once again to Fram's original questions.

1. DannyH (WMF) has made it adequately clear that WMF has no intention of respecting the consensus of the RFC. That question we can consider defitively answered.
2. The second part of the question does not appear to have been answered at all, (unless I have somehow overlooked it), which is when we may expect the magic word to be available? · · · Peter (Southwood) ^(talk): 08:10, 16 February 2018 (UTC)[reply]

For question #1: The one thing that we're asking for in this compromise magic-word solution is that the existing descriptions don't get immediately mass-blanked, breaking a useful feature for the readers and editors that use it. The RfC consensus is that the descriptions should be immediately mass-blanked. We're not willing to take a useful feature away from people who use it over an issue that is really just about the timing of stage two. Question #2: I haven't checked in this week with the engineer who's working on this -- I'll ask him, and I'll let you know what he says later today. -- DannyH (WMF) (talk) 15:39, 16 February 2018 (UTC)[reply]

Okay, I heard back from the developer -- he estimated that it'll be about two weeks before the magic word is deployed. It's always possible for something (related or unrelated) to come up, so that's an estimate and not a promise. But it'll be in that area, and if any delay comes up, I'll post about it here. -- DannyH (WMF) (talk) 20:47, 16 February 2018 (UTC)[reply]

Thanks DannyH (WMF). That seems a reasonable timeline. Cheers, · · · Peter (Southwood) ^(talk): 14:22, 17 February 2018 (UTC)[reply]

• I remain furious about this. Volunteer time is the lifeblood of WP. Because of the clueless arrogance of the WMF reading team, volunteers now are starting a "major undertaking" to fix their damage - doing that instead of building and updating content. And these short descriptions will inevitably lead to disputes about precisely what to say, which will be yet a further waste of time.

The WMF reading team should never have intervened into content this way in the first place. Clueless and arrogant breach of the basic deal here - the editing community generates content; the WMF just publishes it. Now the tail is completely wagging the dog.

I recognize that the community has basically accepted this power play/blackmail via the RfC, but it is horrible. Because of the RfC I will not remove these tags, but I am going to ignore this except to remind everybody from time to time that our work has been hijacked, and that this is a capitulation to an exercise of power by our publisher over the editing community. Raw power. Ugly stuff. Jytdog (talk) 16:27, 17 February 2018 (UTC)[reply]

On a more constructive aspect, I have opened a discussion at Wikipedia_talk:Manual_of_Style#Article_short_descriptions, to consider what we want the short descriptions to look like. · · · Peter (Southwood) ^(talk): 14:55, 8 February 2018 (UTC)[reply]

I also started a page Wikipedia:Short description to direct people to who don't know what is going on when I add short descriptions. I am using the temporary template {{short description}} for now which can be replaced later by the magic word to save some time, as crafting a good short desription is not always a trivial exercise. · · · Peter (Southwood) ^(talk): 16:12, 14 February 2018 (UTC)[reply]

DannyH (WMF), Would it hinder the developers in any way if we set up a temporary {{SHORTDESC}} on Wikipedia so that we can start populating with the short descriptions? · · · Peter (Southwood) ^(talk): 15:38, 8 February 2018 (UTC)[reply]

There is a fair chance that the syntax will be different: template takes {{SHORTDESC|desc}}, while magic word probably takes {{SHORTDESC:desc}} (a colon instead of a pipe). So any descriptions you would set up would need to be changed anyway. Fram (talk) 15:45, 8 February 2018 (UTC)[reply]

That is a pain, but brings me to another question.

Is there a semi-automated method by which we can work through a set of articles, which would check if there is already a short description, and if not, open the article to edit, suggest the Wikidata short description, and add it on a click of acceptance, otherwise add the magic word syntax in the right place, wait for the user to type it in and save, then move on to the next article? Such a tool could accelerate the process by an order of magnitude. · · · Peter (Southwood) ^(talk): 16:07, 8 February 2018 (UTC)[reply]

Any bot could do it. This was actually what I proposed at the RfC. If you really want a click in the end (I am not sure why, there are too many articles), probably a wizard like Magnus Manske can write it (it would be similar to a Reasonator). I am not sure though people at the Wikidata side would be super happy to write tools which result from a clear rejection of Wikidata.--Ymblanter (talk) 16:15, 8 February 2018 (UTC)[reply]

Oh I don't know, Ymblanter. We're going to get human eyeballs on a lot of descriptions out of this, and with luck a pile of improved descriptions that can be imported straight back to Wikidata. That's quite an upside!

A few thoughts

• What Peter proposes above isn't far off what can be done with AWB, though I've not had that much experience with AWB, and I've never quite worked out if there's a way to give it a list of a articles and a list of strings, if the string to be added is different for each article. A bespoke tool would be a lot more user-friendly however, and (though I'm not a tool-maker) I wouldn't have thought would be too hard to create.
• Nevertheless, 5.5 million is a lot of descriptions to create/review. I would think, as Ymblanter suggests, that some heavier automation would be needed, ie auto-writing some sort of draft description to pages.
• To highlight what's going on, if a description is added to a page, IMO it would be good to also add a section on a talk page, saying what has been added, how desktop people can see it and edit it, perhaps with a templated checkbox that it had been reviewed and considered adequate.
• Such a template is likely also to kick off a healthy discussion process on the more heavily curated pages over the precise text to show. This would be no bad thing, but we should be cautious not to move faster than WikiProjects and editors seeing things across their watchlists can cope with. For individual WikiProjects, the initial set of descriptions added may set off intense discussion, that it would be good not to flood; but after that, once the descriptions have started to become more of a known territory, it should be possible to accelerate.
• Putting a template on each talkpage saying "this is how you change the short description" will be like painting up a target; it will make these descriptions much more high-profile than they have been up until now on Wikidata. It will be important to make sure that anti-vandalism assessment code is ready for this, ie primed to pay special attention to changes to short descriptions, in particular e.g. changes that seem to produce changes that bear no relation to the properties of the corresponding item on Wikidata. Such assessment software will need to have a good idea of the range of what good descriptions look like; and probably will also take time to settle in. Jheald (talk) 16:42, 8 February 2018 (UTC)[reply]
• One further thing, as I wrote on the MoS thread: does anyone have a link to the code that currently produces auto-descriptions when there is no manual description on Wikidata? Looking at this may give a useful typology of cases to consider, and a useful set of initial formats to critique. Jheald (talk) 16:48, 8 February 2018 (UTC)[reply]

• See d:Wikidata:Automating descriptions for a link to Magnus Manske's AutoDesc. This picks up information from the statements in the item entry. StarryGrandma (talk) 19:54, 8 February 2018 (UTC)[reply]
• Jheald, Some good ideas there. The reason I would prefer that the process be semi-automated is that the Wikidata descriptions are very inconsistent in quality, and there are a lot of them that we dont want to import. After all, that is the reason for this whole fiasco. Semi-automated will be slower, but will encourage relatively fast improvement. There will be an ongoing task on WP to add short descriptions to new articles and fix/improve them on old articles, so worth investing some coding time to make a really good tool. One of the functions will be adding a category indicating the presence of the SD, which will help find the pages without them. Talk page notices are a particularly good idea. A short notification of when the description was added and the name of the editor would encourage good practice and due diligence, and make the change known to those not watching the article and who choose not to make the SD visible on desktop view. As to Wikidata copying them back, sure they are free to do it if the new SDs suit them better than the old ones. This may well be most of the time, but probably not all of the time. Many of the unsuitable for WP descriptions look adequate for WD at present. It may even be possible to code the tool to do this function as well, which would be a gain for both projects.· · · Peter (Southwood) ^(talk): 04:46, 9 February 2018 (UTC)[reply]

Getting back to my original question about the possibility of a template. I assume the magic word will do the things that WMF wants it to do, not necessarily the things we want it to do, so there may be some value in a template after all. One which calls the magic word as one of its functions and passes over the short description character string as a variable. Other functions would include adding maintenance categories, and possibly other things. Or am I over-complicating this? · · · Peter (Southwood) ^(talk): 22:34, 9 February 2018 (UTC)[reply]

I'm not sure that's how magic words work -- I'm not sure you can template them. I think it may be that they have to be in the raw wikitext of the page. But it's not something I have ever looked into.

The advantage of a template is that it would be very easy for a bot to convert into a magic word when the time came. The disadvantage is -- would anyone have any interest in it, until it did actually do anything?

If we want to start getting descriptions onto pages before the magic word is available, it's almost the only game in town. And of course it can be used to set categories. A downside is that it is hard to query at scale -- if you want to review a list of the descriptions on a particular set of articles, I think you'd pretty much have to write a scraper -- I don't think there's any other way to get the values of a template for a whole set of pages. That said, {{GeoGroup}} seems somehow to manage to do it, but that may be a very special case.

One alternative, that would allow querying, would be to ask for a new property on Wikidata, specifically for short descriptions in development. But in the current climate, I can see that's not likely to be a runner.

My recommendation then, until the magic word is available, would be to have a new template, but to put it on the talk page, where one could make it fully visible, where it would trigger watchlists, and likely be a focus for discussion. The only thing one would need to do would be to make very sure it didn't get archived by archive bots.

Are you on good terms with any wikiprojects that might consider trialling such a thing? Jheald (talk) 23:16, 9 February 2018 (UTC)[reply]

Jheald, As the most active editor on WP:SCUBA I have been testing it there, I have written about 477 over 600 short descriptions using a template {{short description}} coded for the exercise by one of the other major WP:SCUBA contributors, RexxS who is skilled with templates and Lua. The template also adds the article to a maintenance category - . There has been some resistance where articles are tagged for several projects, but this has been minor, and I just leave them for later. The main objection has been that it is experimental and therefore not necessary, and so they can delete it. Less than 1% reversion rate at a guess. At present there is no display on the read view, but it could be used to test display formats.

I am also mildly and tangentially active on WP:MED, as there are some diving articles that are about medical topics, and on WP:OSH, for similar reasons. WP:MED is big and active and high profile, and would probably be very serious about the quality of short descriptions. Some of their more active members are aware of the short description saga and may be sympathetic to trials. Biographies are outside my field of interest and competence, but also a project where the quality of short description would likely be an issue. · · · Peter (Southwood) ^(talk): 13:02, 10 February 2018 (UTC)[reply]

Peter (Southwood), I strongly agree that the magicword should be embedded in a template rather than used directly in articles. The template method opens a wide range of valuable functionality and flexibility that is is unavailable in a bare magicword. Alsee (talk) 12:11, 14 February 2018 (UTC)[reply]

Alsee, Besides managing maintenance categories, do you have any suggestions for what else such a template could or should do? · · · Peter (Southwood) ^(talk): 20:32, 14 February 2018 (UTC)[reply]

Peter (Southwood) the WMF has been editwarring on Phabricator to deny the community-consensus requested task.[1]

As I mentioned before, one of the benefits of using a template is how flexible it is. Any change to how the magicword works is a big slow process. The WMF has to reprogram it and go through a software deployment cycle. Changes to the template are quick and easy. For example if one felt the magicword was a friendly addition, one might be motivated to build a message that only appears on preview asking editors to helpfully fill in a description. On the other hand if one felt the magicword were a hostile and pathetically-lame attempt to defeat consensus, one might be motivated to implement consensus anyway. A template can translate "blank" in an essentially infinite number of ways. 100% of those ways defeat the WMF's anti-blank filter. We can change the meaning of "blank" far faster than the WMF can roll out a software deployment re-writing the filter. Alsee (talk) 22:52, 14 February 2018 (UTC)[reply]

Alsee, In what way would this have a benefit to anyone? It looks more of a lose-lose scenario. It would also be theoretically possible for Wikipedians to overwhelm Wikidata and delete the short descriptions we don't like, but that is also a lose-lose scenario. This magic word could be a win-win if we all let it, and by all I specifically include the developers. I don't think the magic word itself is an attempt to defeat consensus, but I do think that the way it is currently specified looks like an attempt at one-upmanship when taken in the context of previous negotiation. · · · Peter (Southwood) ^(talk): 06:06, 15 February 2018 (UTC)[reply]

Peter (Southwood) maybe there was some miscommunication? I don't see how implementing the RFC consensus is 'lose-lose'? The consensus is that Wikidata descriptions should not be used for this purpose, and that we should start to start creating local descriptions. There would be two steps to in-effect implement consensus. #1 A bot to ensure the template exists on all pages. #2 If the description template has not been filled in yet, then the template can pass any value we chose into the magicword to suppress Wikidata. If there's no local description yet, it will display as either literally blank or functionally equivalent to blank. Alsee (talk) 04:22, 16 February 2018 (UTC)[reply]

Alsee, It is possible that I misunderstood. I saw this suggestion as the sort of thing that WMF would react to by switching off the magic word and going back to Wikidata for everything, or at this stage, failing to implement it altogether or crippling it even worse. Niether of these outcomes helps us. As far as I can see a Wikipedia consensus can be enforced internally on Wikipedia, but if WMF personnel choose to ignore it there is no direct way to force them to comply short of going over the heads of the recalcitrant parties to upper management, the ombudspersons or the board. You are free to try any of these if you think they may work, and I would support a complaint. This may fall under governance issues and it would be interesting to see what the board would decide. Doc James may wish to comment on this. Starting what is the equivalent to a war of attrition seems likely to distract editors from the more useful activity of getting the short descriptions written and seems a bit pointy.

We can already start adding short descriptions using the experimental template. If we keep the template and embed the magic word then transition becomes very easy, if we need to change it then a bot run can deal with a consistent implementation, which will also be relatively straightforward. The big task is creating good short descriptions, which is not always trivial.

Thirdly, I do not know how a template could be used to force any value we choose into the magic word to bypass the interpretation of a blank parameter as a call to Wikidata. This is simply because I do not have the technical background to model the process. · · · Peter (Southwood) ^(talk): 05:45, 16 February 2018 (UTC)[reply]

User:Pbsouthwood will having the ability to ONLY watchlist the short descriptions in English change anything for you? I would be much more inclined to accept Wikidata's short desciptions when this occurs. Doc James (talk · contribs · email) 08:49, 16 February 2018 (UTC)[reply]

Hi Doc, Being able to watchlist only short descriptions would be helpful for the transition period, but will make little difference to me personally, as I intend to make sure that all the articles on my Wikipedia watchlist have a short description in Wikipedia within a short time, probably the majority even before the magic word is implemented, after which the Wikidata versions will be irrelevant. When I work on Wikidata I don't need that function either. So not for me, but maybe for others it will help get the short descriptions populated on Wikipedia, which would be a very good thing. Cheers, · · · Peter (Southwood) ^(talk): 10:02, 16 February 2018 (UTC)[reply]

Hi Doc James, we've been talking to the folks at Wikidata about getting only the short descriptions to appear in English watchlists. They've been working on this for a while, but unfortunately, there isn't a current plan for when this could actually be done. We started working on the magic word override because we don't want this situation to continue indefinitely. -- DannyH (WMF) (talk) 15:54, 16 February 2018 (UTC)[reply]

User:DannyH (WMF) thanks for the update. Yes realize that this is being worked on. I view it as critical for improved collaboration between WP and WD so hopefully it will be done eventually (not just for the short descriptions but for whatever is used on WP). Realize that these things of course take time. The magic word override I view as a less preferable option (but also realize that others feel differently). Best Doc James (talk · contribs · email) 00:36, 17 February 2018 (UTC)[reply]

Doc James, I am not against collaboration with Wikidata, but material from WD that is to be used on en:WP must have quality assurance at least approximately equivalent to en:WP in practice. At present they do not even have compatible policy, which is a total dealbreaker for some kinds of data. Compared to getting Wikidata policy and QA aligned, which they may not want to do, having the short description on Wikipedia is an easy solution. · · · Peter (Southwood) ^(talk): 05:59, 17 February 2018 (UTC)[reply]

Peter (Southwood), first regarding the template: My template knowledge is partial, but I created a sandbox version[2] to illustrate what I mean. (I think there's a simpler way to do it, but what I wrote is working.) If you view the testcases page[3] with that sandbox version live, you should see that it passed any description through to the magicword. However if no description is provided, it passes the value "NONE" to the magicword instead. That would suppress Wikidata and display "NONE" as the description. NOTE: I only used "NONE" as an illustration. I have better ideas in mind, which may or may not generate a literal blank. I am deliberately avoiding being more specific. If the WMF expands the filter to block the best values, it merely means we'd use non-best alternatives. No filter can stop us from passing a suppress-wikidata-value, short of manually reviewing each value.

Regarding what we can and cannot do, you're right that we can't force the WMF to build anything. However you underestimate what we can do, and the WMF somehow still keeps underestimating what we can do. There have been many instances in the past where the WMF presumed they could ignore the community, presumed the community couldn't do anything, because something is supposedly "serverside". Somehow the WMF is still surprised when, time after time, the community demonstrates it can in fact modify or disable things from our end. If the WMF insists on battling the community they will almost certainly end up in a worse position than if they had just build the consensus-fix in the first place. And as you note, the WMF efforts to pit Wikidata against Wikipedia may very easily escalate or detonate the existing powderkeg of inter-community tensions. Alsee (talk) 08:04, 16 February 2018 (UTC)[reply]

Hi Alsee, I looked at your template edit and see what you are getting at. There are a very large number of ways to say There really should not be a short description here but WMF is forcing us to provide one so we have provided this little message to be pointy about it. This would not stop WMF from shutting down the magic word and going back to all Wikidata, which would be moving in the wrong direction and might ignite the powder keg. I would prefer not to have any inter-community detonations, the fallout would affect everyone, and while I am primarily a Wikipedian, I also work on other projects, including Wikidata, and have some sympathy for them, as they are at risk of losing the most. Disabling the magic word would also just mean WMF going back to Wikidata, which would be a very poor result. I entirely agree that building the consensus fix would save a lot of pain and possibly jobs, but I don't know the internal politics behind this, so can't predict what risks they would be prepared to take to force this through. · · · Peter (Southwood) ^(talk): 10:02, 16 February 2018 (UTC)[reply]

Please see Wikipedia talk:Manual of Style#Article short descriptions. While it's on the MoS talk page, it's really a continuation of this discussion, and is primarily about a) the idea of using an en.wp template; b) whether using WD-provided descriptions should be done at all, and if so how/why/when/to what extent/etc.; and c) the inherent difficulties of using descriptions as short as 40 chars. Personally, I think this discussion should be centralized, probably here not at WT:MOS (it has virtually nothing to do with style, but with content), and then particular things eventually broken out for further practical refinement at WP:VPTECH. At some point these ideas have to get more buy-in from the community than just the people who talk about WD all the time, and specific solutions need to be put into place, probably requiring some Phabricator tickets.  — SMcCandlish ☏ ¢ 😼  23:51, 9 February 2018 (UTC)[reply]

I notice that the phabricator ticket specifically by-passes the ability for Wikipedians to specify a blank short description.

If the magic word isn't used on the page, or if the description is blank, then it should show the description from Wikidata.
"Blank" means:
not having any description in the field {{SHORTDESC:}}
just blank spaces {{SHORTDESC: }}
just punctuation {{SHORTDESC:.}}
just non-breaking space {{SHORTDESC:<code>&nbsp</code>;}}
In those cases, it should pull the description from Wikidata.

I consider this an act of bad faith and a clear disregard for the consensus on the part of DannyH (WMF) until a satisfactory explanation is forthcoming. · · · Peter (Southwood) ^(talk): 06:40, 11 February 2018 (UTC)[reply]

In stage one of this process, if {{SHORTDESC: }} blanked the description, then it would be easy for someone to run a bot and put that on every page. I have said consistently that mass blanking of the descriptions is not going to work for us. Once there are ~2 million descriptions on English WP articles, we move to stage two, and at that point any page that doesn't have a SHORTDESC magic word will automatically be blank. This is the same thing I've said throughout the discussion.

Regarding the consensus, I had offered to co-write the RfC with Alsee, so that our point of view would be included. Alsee didn't start writing it, and then Fram went ahead and created the RfC without my input. Still, I participated actively in the conversation, and I expected that my point of view would be included in the consensus. A month into the RfC conversation, I came up with a new proposal for a compromise -- the one that's on that Phabricator ticket. This was belatedly added to the RfC choices as #5, and you (Pbsouthwood) crossed out your original statement and switched to #5 as your preferred choice. Other people didn't change their original statements -- this was a month after the start, and most people had drifted away -- so it wasn't really considered in the final consensus summary. Still, I think this #5 compromise is reasonable and fair, and so did you, as of January 6th. -- DannyH (WMF) (talk) 21:29, 11 February 2018 (UTC)[reply]

@DannyH (WMF): What about disambiguation pages, list pages, category pages, and other pages where the content may be implicit in the name? Jheald (talk) 22:32, 11 February 2018 (UTC)[reply]

Jheald, I was thinking that would be addressed after people had written the article page descriptions and we switch over to just using Wikipedia-hosted descriptions. At that point, all of the category, list and disambig descriptions would disappear, along with all of the other unwanted descriptions. -- DannyH (WMF) (talk) 16:54, 12 February 2018 (UTC)[reply]

@DannyH (WMF): Well, as part of the descriptions creation process, I think it would be good to be able to indicate affirmatively the cases in which no description is the preferred choice, and for that then to take effect there and then. Is there a problem with that? Jheald (talk) 17:04, 12 February 2018 (UTC)[reply]

DannyH (WMF), your understanding of the Wikipedia community and culture is deplorably lacking in someone who is acting as a spokesperson for WMF in negotiations with Wikipedia if you do not understand the workings of consensus on Wikipedia. The fact that I think that option 5 was a reasonable compromise does not override the fact that the RFC was closed as consensus for option 2. If you think the closure was wrong (the closure was reasonable by my assessment, but not indisputably so), you can reopen the RFC, you do not say nothing and immediately proceed blatantly against the consensus. That is the kind of thing that can get you sanctioned on WP, which would not help your work at WMF.

You made your points at the RFC, they were mainly rejected. I do not think that if you had co-written the original question it would have greatly affected the outcome. Better, more timeous replies, and less ignoring the questions and giving partial answers apparently avoiding the issues might also have helped.

Your counter-consensus specification on the phabricator ticket goes beyond simple disagreement into the realm of actively obstructing the preference of Wikipedians to be able to leave out a short description in cases where Wikipedians agree that a short description is unnecessary or even undesirable, as Jheald mentions above.

It also appears that there is a misunderstanding regarding the use of short descriptions. Wikipedians were expecting (to the best of my understanding) that all mainspace articles will be populated with the magic word, not that the magic word would be omitted from some articles to later display no short description. That is an arrangement that will cause endless confusion, as editors will be adding and removing the thing at great waste of time and effort. A magic word present in the article, with a comment to the effect the there is local consensus to the specific short description avoids that confusion and allows easier maintenance categorisation. This you just dismiss without discussion, and do as you please. Not a way to instill confidence in WMF within the community Again one is reminded of superprotect.

You appear to misunderstand the process of bot runs on Wikipedia. Anyone who ran a bot to install blank descriptions across the whole Wikipedia without consensus and specific permission, would immediately face sanctions, and most likely permanantly lose bot operator permission, and the run would be reversed. I doubt that we have that many suicide botters.

There is more I could add, but the wall of text is already uncomfortably high. · · · Peter (Southwood) ^(talk): 06:34, 12 February 2018 (UTC)[reply]

Pbsouthwood, we've been talking about this for a long time, and at the end of the day, you and everybody else on English Wikipedia are getting exactly what you want. The engineers are working on the magic word, which will allow editors on WP to write the short descriptions. Once that's deployed and there's a roughly comparable number of Wikipedia-hosted descriptions, we'll switch over to entirely Wikipedia-hosted descriptions. This is a simple question of timing, and that timing hinges entirely on English WP contributors doing the thing that you're currently working on doing. -- DannyH (WMF) (talk) 16:54, 12 February 2018 (UTC)[reply]

DannyH (WMF), If we do get approximately what we need at the end of this it will only be as a result of constant vigilance and seeking out the deviations and protesting them in public. I will once more appeal to logic and common sense and request that the counter-consensual specification for bypassing a blank magic word to Wikidata be removed from the specification. It is against the requirements of English Wikipedia, and though the added complexity may be small, it is an added complexity to make those tests. We do not want this added dysfunctionality and the developers will have less to code and test without it. You personally will be less resented on Wikipedia, so we all win. Cheers, · · · Peter (Southwood) ^(talk): 17:13, 12 February 2018 (UTC)[reply]

Pbsouthwood and Jheald, what's the plan for how the magic word is going to be populated? One way to do it would be to use that bot to add the {{SHORTDESC:}} to every page, which will be a call-out for people to start filling them in. But if that's how it's done, then honoring those blank descriptions would mean a complete blanking of all descriptions for the users.

I've said many times, before, during and after the RfC, the one outcome that I need to be assured of is that the descriptions should not be mass blanked. The consensus specifically says that is the outcome that you want -- "start with blanks, allowing to fill in manually" and "show no description for blanks". I don't know how I can make it more clear that that the WMF is not going to build that. If that is the thing that you're asking for right now -- to immediately mass-blank every description before people start to fill them in -- then I'm not sure how to make this conversation more productive. -- DannyH (WMF) (talk) 05:26, 13 February 2018 (UTC)[reply]

I was under the impression that WMF would only turn the magic word on when they (you?) arbitrarily decided (Quoting a number of 2 million, and whether or not this was agreed by Wikipedians) that there were enough short descriptions on Wikipedia, not that the magic word would be switched on immediately before we even start adding short descriptions. If this was not your intention, you could try to communicate more clearly. On the reasonable assumption that you would be doing what you said you would, mass population with blank descriptions would have literally no effect on the search display function until the magic word is activated, which you said would only happen when we had put in about 2 million short descriptions, based on your statistically dubious estimate of the numbers on Wikidata. We regard this as a form of extortion, but it would require a major campaign to fight it, which may not be worth the trouble.

Some of us are not keen for inappropriate short descriptions to be copied over from WikiData, others may not care. This probably means that bot runs requests will be watched and debated if there is any likely controversy, so reducing the risk of mass blank descriptions anyway. I am advocating semi-automated runs, requiring a user check on each short description copied or generated automatically from whatever source the user chooses, in the interests of quality assurance. This is still early days, and I cannot predict the outcome.

Whatever bots, if any, are used to add the magic word to articles, they will have to go through bot approval for each run first, We do not allow bots to run wild at the whim of any lunatic.

This conversation would probably be more productive if there was a clear description of what is planned by WMF, not a vague (mis)understanding based on several things stated in various places at different times and subject to personal interpretation and arbitrary movements of the goalposts. Since it is WMF who caused the problem, and will be writing the code to fix it, it would be appropriate if the person running the WMF side would write a clear and unambiguous explanation of what WMF is planning to do, preferably with a timeline, in language that the average Wikipedian can understand. · · · Peter (Southwood) ^(talk): 06:29, 13 February 2018 (UTC)[reply]

Pbsouthwood, unfortunately, you have misunderstood this. I'll explain it more carefully.

The magic word is currently being built. When it's complete, it will be deployed to Wikipedia and will be live. There are two stages after the magic word goes live:

Stage 1: Wikipedia editors will populate the magic word (SHORTDESC) on Wikipedia pages. During that period:

• Pages that have a Wikipedia-written SHORTDESC description -- {{SHORTDESC:American stage actor}} -- will display the new description.
• Pages that have a blank magic word -- {{SHORTDESC:}} -- will display the Wikidata description.
• Pages that don't have a SHORTDESC description will display the Wikidata description.

Stage 2: Once Wikipedia editors write ~2 million descriptions, we'll switch to entirely Wikipedia-hosted descriptions. From that point:

• Pages that have a Wikipedia-written SHORTDESC description -- {{SHORTDESC:American stage actor}} -- will display the new description.
• Pages that have a blank magic word -- {{SHORTDESC:}} -- will not display a description at all.
• Pages that don't have a SHORTDESC description will not display a description at all.
• The Wikidata description will not be displayed on any page.

This is the same plan that I posted in the RfC on December 22 under "WMF two-stage proposal for Wikipedia-hosted descriptions", and it's the plan described in the Phabricator ticket.

What we've been talking about for the last couple days is what happens in Stage 1 when a page has a blank magic word. I hope that this explanation clears things up. What do you think? -- DannyH (WMF) (talk) 19:39, 13 February 2018 (UTC)[reply]

@DannyH (WMF): It seems a clear enough statement of your position. I'm not sure Pbsouthwood will be very happy with it, as there is an element of WMF over-ruling the decision of an en-wiki RFC. So we'll see how that goes.

Specifically on the question of null descriptions, however, I do think it would be useful to be add these in-process as the process goes forward. IMO it would be a lot better if the results were immediate, as that is the time when people will be actively looking at the question. So, if an explicit directive were added for no description, eg {{SHORTDESC:NODESC}}, is there any objection from you in principle to honouring that straight away, albeit perhaps with you reserving the right to cease honouring it were you (or whoever would be WMF's relevant assignee) to consider it was being significantly misused? Jheald (talk) 00:49, 14 February 2018 (UTC)[reply]

At first reading, and without prejudice, this appears to be a clear and unambiguous description of the magic word and implementation process such as I have been requesting for a long time, probably months. It might have saved a lot of hassle if this had been provided earlier, possibly even when first requested. I do not remember this information being made available before, and I have been watching out for it, as I asked for it.

I will go back to the statements of December 22 in the RFC before commenting on whether this describes a plan previously posted, but it does not look familiar in this format. It looks very unfamiliar.

Similarly for the plan described in the Phabricator ticket.

At this stage, I mostly agree with both of Jheald's comments directly above. More later. · · · Peter (Southwood) ^(talk): 05:55, 14 February 2018 (UTC)[reply]

I have found a posting of December 22 that I assume is the one you refer to

* Stage 1: Initially, the display of the descriptions pulls from the Wikipedia-hosted magic words -- but if there isn't one on Wikipedia, or the Wikipedia one is some version of blank, then it falls back to showing the Wikidata-hosted description.
* Stage 2: When there are non-blank Wikipedia-hosted magic words on a number of articles that's roughly comparable to 2.88 million, then WMF switches to only pulling from the Wikipedia-hosted magic words, and we don't fall back to Wikidata-hosted descriptions. At that point, descriptions that Wikipedia editors leave blank will actually be blank on the site.

This essentially the same as your clarified explanation above. This was rejected by consensus at the RfC. Basically what you are saying appears to be that you will override the requirements of the Wikipedia community if they do not fit in with your own requirements and that Wikipedia policy is subordinate to the convenience or preferences of a WMF department as determined by yourself and unnamed others. · · · Peter (Southwood) ^(talk): 07:34, 16 February 2018 (UTC)[reply]

I think you are confused about who WMF answers to. They answer to the board, not the community. If any of the above has you surprised then it's time you get more involved with board decisions. A single wiki's consensus is not more than feedback for the WMF, it has zero meaning whatsoever outside that single wiki. —TheDJ (talk • contribs) 11:12, 16 February 2018 (UTC)[reply]

I've made it very clear since this conversation started months ago that sudden mass blanking of the short descriptions is unacceptable for the WMF, because it would break a feature that readers and editors find useful. The compromise solution of the two-stage magic word gives Wikipedia editors the ability to override and replace all of the Wikidata-hosted descriptions. The only thing that the WMF is asking for is to not mass blank all the descriptions while the Wikipedia editors work on populating the descriptions, so that the readers and editors who use the descriptions don't suddenly lose a useful feature. Despite this offer of a good-faith compromise, the RfC consensus said that we have to mass blank all descriptions immediately.

Rather than escalate tensions, I think it would be a better use of everyone's time to figure out how to add descriptions to the pages. If the break from Wikidata-hosted descriptions needs to happen immediately when the magic word is created, then one possibility is to run a bot that imports the existing descriptions from Wikidata. Then Wikipedia editors can replace these with Wikipedia-created descriptions at their own pace. -- DannyH (WMF) (talk) 15:32, 16 February 2018 (UTC)[reply]

TheDJ, I am aware that they are answerable to the board, but I am not aware of any mention of the board having been consulted in any way during these discussions. In view of previous fiascos I would have thought it prudent to get their opinion at some time during the months that this had been dragged out over.

DannyH (WMF), Can you assure us that if we do import the Wikidata descriptions wholesale, that the break will be made immediately on implementation, and not reverted even if we find it necessary to delete a large number of the imported descriptions as unsuitable? I would prefer to see them improved or corrected, but in some cases deletion may be the most appropriate action. This would also remove the necessity to code in the blank evasion filters.

It would also help if you would answer questions relating to figuring out how to add descriptions to the pages. I asked earlier whether the magic word could be used from inside a template, but the only answers I got were from fellow Wikipedians who basically admitted that they were guessing. If we can add a template which adds a maintenance category and put short descriptions into it, then we can add the magic word to the template when it is ready. Some say magic words don't work that way, others seem to think they can. If you will tell us one way or the other we can get on with something useful. · · · Peter (Southwood) ^(talk): 16:30, 16 February 2018 (UTC)[reply]

Pbsouthwood, I assume everybody will be acting in good faith. When there are roughly ~2 million descriptions on article pages and we switch off the Wikidata fallback, then the content is under Wikipedia contributors' control. There are a lot of descriptions coming from Wikidata that I agree should be blank -- for example, the list pages and disambiguation pages, as well as other examples that I'm sure people will identify in the future. The only situation that we want to prevent is the sudden, immediate mass blanking of all descriptions; that's why we want the two-stage process. The "blank evasion filters" are only meant to make the transition from Wikidata-hosted descriptions to Wikipedia-hosted descriptions smoother for the readers and editors that depend on the descriptions. When we switch to stage 2, all the content decisions are made by Wikipedia editors.

Sorry I didn't answer the question about templates -- I didn't answer because I don't know, and I'm not sure who does. My main experience with magic words is with DEFAULTSORT, which is a free-standing magic word usually placed at the bottom of the article's wikitext. The code for displaying the descriptions on article pages will be pulling from the magic word that's written in this form: {{SHORTDESC:Novel by Charles Dickens, published in 1859}}. The description display doesn't appear in the page itself. I've read the conversations about using it in a template, and I have to admit I don't understand what people think that the template would do. That's why I've been staying out of those conversations. :) Does anyone know an example of an existing magic word that sits inside a template, and what it does? -- DannyH (WMF) (talk) 18:15, 16 February 2018 (UTC)[reply]

Well that came as a bit of a surprise. You have people coding a magic word but they can't tell you if it will work from in a template? The reason we want to put it in a template is twofold. Firstly, with a template we can automatically add suitable categories, which are useful for maintenance. With a category we can easily check which articles have a short description and which do not, without a category it is much more difficult to find the pages which still need one. Secondly it should be possible to get it to display where and how we want it by adjusting the template, without having to add other code to each page as we will want to be able to display it for editors to inspect on the regular reading display in desktop view as an option. That visibility helps as people who can see it are more likely to fix it if there is a problem. Do you think you could also ask your coding team how we can extract the short description on other pages for on-wiki purposes, We also want to use it for other projects where we can display it in association with a link to the page in lists, where the short description display can be toggled on or off. We would like to be able to call a display of link and short description from a template. For example {{listing|Joe Bloggs}} might display as [[Joe Blogss]] – short description on the Joe Bloggs article · · · Peter (Southwood) ^(talk): 19:47, 16 February 2018 (UTC)[reply]

Pbsouthwood: The feature that we're building is a magic word that will do what I described -- an override for the Wikidata description that will appear in the places where the Wikidata descriptions currently appear. For example, in the Wikipedia iOS app, the short descriptions appear in the "top read" module in the feed, as you can see in the picture to the right. Currently, the code that displays those descriptions pulls the short description from the Wikidata entry. Once the new magic word is deployed in a few weeks, that code will display the short description from the {{SHORTDESC:description}} on the English Wikipedia article. That's all the magic word does -- replace the short description as it currently appears in the places where it appears.

You want to start displaying the short descriptions in new places, including desktop view. My understanding is that other Wikipedia editors have been asking us to not display the descriptions in desktop view (which we haven't done). The maintenance categories, extractions and toggle displays are all new feature requests that aren't directly related to the feature that we're building. As I said, the purpose of the magic word is to override (in stage 1) and replace (in stage 2) the existing descriptions in the places where they currently appear. Any templates, gadgets or tools that Wikipedia contributors build around that magic word is up to them/you. -- DannyH (WMF) (talk) 21:07, 16 February 2018 (UTC)[reply]

DannyH (WMF), I understand the limitations of the coding being done, and that anything we do with the magic word is up to us. What I am trying to clarify is whether it will be actually possible to use your product in a way that is compatible with the needs of Wikipedia and the way things are done here. As neither of us seem to have any idea of how this magic word can be integrated with the template system which we use for so many maintenance and display functions, this is proving to be a case of the blind leading the blind. If we cannot add categories and display the short descriptions by optional scripts, it may be necessary to add a whole lot of overhead to the articles (every one of them) that may affect server load significantly (or not, I really don't have a clue). I am trying to get clarification at a stage where it is possible to make this work efficiently, and not end up needing a whole series of hacky fixes for unforeseen consequences. If your department fails to do due diligence, we have to live with the consequences. What we need to know is that your code for displaying the short description from the {{SHORTDESC:description}} on the English Wikipedia article will be able to find the code and extract the short description if it has been transcluded in a template. If it can do that, then we think we can make it work. At present it is our plan to transclude it in a template because that will probably do all the things we need. If that will not work we need to know, so we can try to work out if there is another way it can be used without breaking the encyclopedia. · · · Peter (Southwood) ^(talk): 06:42, 17 February 2018 (UTC)[reply]

Pbsouthwood, I asked a colleague about this, and here is his reply: "I think what they're saying is you could have a template like {{page description|This is the description}} so in the template code, it'd look for $1 (the "This is the description") part, and if it's not there, put it in a category. That would imply we use {{page description}} on every page, though. Not sure how they are going to accomplish that (no way to make a template transclude on a page automatically, I don't *think*). But to answer the question, yes I think you'd probably be able to use SHORTDESC in a template, just like it is done sometimes with other magic words, e.g. Template:DEFAULTSORT." -- DannyH (WMF) (talk) 20:09, 17 February 2018 (UTC)[reply]

Thanks DannyH (WMF), We would put the template on the page at the same time that we put the short description there. In some cases it might be automated by a bot run, such as for disambiguation pages, other times it might be semiautomated, using AWB or a script to copy useful short descriptions from Wikidata after checking that they are suitable, or manually, when there is no better way available. The template would allow us to easily find which pages still need a short description, by putting those with a short description into a category, so would be a critical part of eventually producing a complete or near complete set of short descriptions. Please check with your programmer who is coding the magic word to ensure that the word will work this way. Cheers, · · · Peter (Southwood) ^(talk): 21:20, 17 February 2018 (UTC)[reply]

Pbsouthwood: Yes, that will work. -- DannyH (WMF) (talk) 20:36, 18 February 2018 (UTC)[reply]

Pbsouthwood if the template has no description value, the template can easily detect that and add a category for missing-description. However I don't know of a good way to track pages that don't have the template at all. Alsee (talk) 21:05, 19 February 2018 (UTC)[reply]

Alsee, Detecting an absence is always a more tedious process. The only way I know of is by checking each article systematically. It would be possible to do a bot run to add Category:Lacks short description, which would then make a cross category search using PetScan almost trivial. Then it would be necessary to remove that category when adding a short description, though that could also be automated as a regular maintenance bot run. This is all adding overhead, but I don't know how much it matters. There may already be a bot doing similar classes of cleanup which could be tasked to do this as well. It would be an ongoing tak as there will always be new articles added without a short description. Cheers · · · Peter (Southwood) ^(talk): 06:24, 20 February 2018 (UTC)[reply]

@Pbsouthwood: Let's not import the Wikidata descriptions wholesale. Many were created with Magnus's autodesc in batch mode, when there was a lot less data in the items. We can probably do better, if we can identify formats that we want for particular sorts of articles (eg default for a album might be <YEAR> <GENRE> album by <CREATOR>, similarly for a film with <DIRECTOR> or for a person might be <NATIONALITY> <OCCUPATION>, <BIRTH YEAR>-<DEATH YEAR>. If the Wikidata descriptions are already of this form, before copying them over it would be good to include any missing information. Of course, hand-created Wikidata descriptions could be recognised as not fitting the pattern, and offered as-is. Jheald (talk) 17:21, 16 February 2018 (UTC)[reply]

Jheald, I am certainly not going to do something like that unless there is community consensus, particularly as I do not run bots myself anyway. I am looking for viable options and useable compromises that will get us what we require in spite of the difficulties. Depending on how long the magic word takes to code and debug, we can partly prepopulate with handcrafted short descriptions, which will not be overwritten later. If there are other methods to create good bulk batches, those can also be done, then if or when patience runs out because it is all going too slowly, as a last resort we could bot populate undescribed articles wholesale from Wikidata to get the Wikdata monkey off our backs permanently, but only if WMF is prepared to accept that there will be an unpredictable amount of garbage we may have to simply delete therafter. Cheers · · · Peter (Southwood) ^(talk): 20:03, 16 February 2018 (UTC)[reply]

@Pbsouthwood: There are other ways to use the Wikidata descriptions than just importing them en-masse unseen. My view is to think that they might be quite useful as draft descriptions. One might eg imagine a tool to show all of the WD descriptions for the pages in a particular category that didn't already have short descriptions, with the ability to either edit them or reject them, and then once one was happy write the edited ones to the pages. Or a tool to offer them, together with Magnus's autodesc, and the first half-sentence of the article, as options, each of which could be selected to copy to an editing line, edited, and then, with an 'approve' click, auto-written into the article.

I think those make sense, as ways to refine the Wikidata descriptions for use here. But doing some automated tidying up on Wikidata first, where sensible, may sometimes be a useful step to help speed the process.

Also, the first version committed of a description isn't necessarily the end result. As I've written elsewhere, I think it would be useful if that would also trigger a message on the talk page about the description, encouraging editors to review it. Jheald (talk) 20:45, 16 February 2018 (UTC)[reply]

Jheald. Tidying up on Wikidata is an option for anyone who wants to do it. preferably before importing, or the effort will be wasted, and providing that the person doing it makes sure that the result is suitable for Wikidata's needs. With a little extra work such enthusiasts can copy the improved short discription over to WP immediately. ( I have done several of these myself, but there is a short but steep learning curve when one starts. I am aware of Wikipedians who consider this an insurmountable obstacle )

A message on the talk page could be useful to notify that an unreviewed short description has been added (on the first occasion). Do you have any idea of how this could be done? Would this only be for bot runs?

You can also make this kind of suggestion at Wikipedia:WikiProject Short descriptions, which you might wish to join. · · · Peter (Southwood) ^(talk): 06:42, 17 February 2018 (UTC)[reply]

@DannyH (WMF): {{DISPLAYTITLE}} is used inside {{italic title}}, {{Lowercase title}}, and some similar templates. It changes the title displayed at the top of the page (with limits on use). {{3x|p}}ery (talk) 01:38, 17 February 2018 (UTC)[reply]

Thanks for the example, Pppery. -- DannyH (WMF) (talk) 04:05, 17 February 2018 (UTC)[reply]

Pppery, thanks for the information. Can you say whether this is a general characteristic of magic words or whether they have to be coded specifically to allow this use? this is important for project maintenance and making the feature more useful. · · · Peter (Southwood) ^(talk): 05:46, 17 February 2018 (UTC)[reply]

I have been able to find a large number of examples of magic words of type=variable that are routinely used inside templates, so I am going to make the reasonable assumption that in the absence of any indication that it will be different, {{SHORTDESC}} will follow this trend, and will set up systems to use it based on this assumption. · · · Peter (Southwood) ^(talk): 13:20, 17 February 2018 (UTC)[reply]

Hi all, work on the magic word is progressing well. The code is currently up for review, and we're expecting that it'll be released next week or the week after. I'll let you know if I find out anything more. -- DannyH (WMF) (talk) 02:07, 7 March 2018 (UTC)[reply]

Hi all, here's the progress. The magic word is sending the override to the API. To test it, I used my volunteer account to add {{SHORTDESC:}} on two article pages this morning (Guy Madison and Wagon Train). The API is registering the override (ex: here). Now we need to update the places where the short description is used, to pull from that API. The work is being documented on this ticket, if you're curious: phab:T184000. I'll post when I learn more. -- DannyH (WMF) (talk) 16:46, 28 March 2018 (UTC)[reply]

DannyH (WMF), How does a logged in editor using desktop view see the content of the magic word, or know that there is one in an article? · · · Peter (Southwood) ^(talk): 20:14, 28 March 2018 (UTC)[reply]

Pbsouthwood: Right now, it's only visible in the wikitext, not in the desktop or mobile web view. Last year, there was an RfC that ended with the consensus to take the short description off of the mobile view, so we haven't worked on showing the descriptions in more places. The descriptions are currently seen in the apps, in Visual Editor, in the Content Translation tool, and in some search results. -- DannyH (WMF) (talk) 21:08, 28 March 2018 (UTC)[reply]

DannyH (WMF), I see that. However, as it is the Wikipedia editors who will have to produce and maintain the short descriptions, it is an obvious functional necessity that we are able to see them, so we can fix the ones which are unsuitable, and identify which articles do not have them, so they can be added. What is the WMF plan to make this possible, and when can we expect it to be running? I assume you do understand that there is unlikely to be any serious objection to the use of Wikipedia based short descriptions in mobile view, as the objection to their use was that they were poor quality and not under the control of Wikipedians, but this would be conditional on Wikipedians actually being able to see and edit them reasonably conveniently: ie. they must be under our control.· · · Peter (Southwood) ^(talk): 06:39, 29 March 2018 (UTC)[reply]

Pbsouthwood, It would probably be best to start an RfC about that. I agree that it's a good idea for editors to see the short description, and the situation has changed since last year's RfC. I think that RfC would be most likely to succeed if it came from editors, rather than the WMF. Can I do anything to help? -- DannyH (WMF) (talk) 16:35, 29 March 2018 (UTC)[reply]

DannyH (WMF), Would that be an RFC about whether editors should be able to see content of a Wikipedia article from the read view on desktop, or an RFC about WMF using content from a Wikipedia article to represent the Wikipedia article? · · · Peter (Southwood) ^(talk): 16:42, 29 March 2018 (UTC)[reply]

Pbsouthwood: The first one. We've already had the RfC about using Wikipedia content to represent the Wikipedia article, the result was the magic word that we're talking about now. The RfC that's required is about displaying the descriptions on desktop and mobile view. Right now, if my team added a display of the descriptions on mobile web, people would be shocked, correctly, that we made an important interface change that a previous RfC specifically asked us not to do. That needs to be a community-generated proposal. If the community consensus is that we should have the descriptions visible on desktop and mobile web, we'll be happy to make that change. -- DannyH (WMF) (talk) 21:57, 29 March 2018 (UTC)[reply]

OK, we will set one up. Alsee, Fram, Jheald, Mike Peel, RexxS, Jytdog, Doc James, anyone else following this page, please make any suggestions and comments about an RFC for displaying the short description in the various views available to editors. This would affect every editor and possibly every reader of English Wikipedia, so should be highly visible, and it would be preferable to set up good questions before opening, so the waters are not muddied too much. If anyone can suggest someone neutral and particularly skilled at setting up an RFC, please invite them in. We need to work out what the options are, and whether they should be universal, opt in, or opt out. · · · Peter (Southwood) ^(talk): 06:26, 30 March 2018 (UTC)[reply]

We don't need an RfC, we need someone to write a gadget which shows the description in desktop view, and it should start as an opt-in option in preferences (if it gets a lot of success, then we can have an RfC to change it to opt-out). Fram (talk) 07:39, 30 March 2018 (UTC)[reply]

Fram, A single line of CSS seems do do the job. This has been available for at least a month and is mentioned at Wikipedia:Short description#Template:Short description. "Any registered editor can show the text of short descriptions by adding this line to their Special:MyPage/common.css:"

.shortdescription { display:block !important; }

You can add text colour, background colour, emphasis etc as you please to make it stand out from other text. I use a light yellow background. RexxS uses red italic text. It is surprisingly simple, because the template was planned for this. I suppose a tick-box option could be added to preferences to do the same thing, but I don't know how to get this done. Do you? Cheers,· · · Peter (Southwood) ^(talk): 09:12, 30 March 2018 (UTC)[reply]

Thanks. That displays the "short description" template though, not (directly) the "shortdesc" magic word. In any case, if we have a stable line of css which works for most people (OS, browsers, ...), then a gadget can be requested at Wikipedia:Village pump (technical). That doesn't need a formal RfC, a short discussion may suffice, unless there is too much opposition. Fram (talk) 09:28, 30 March 2018 (UTC)[reply]

I see your point, but are we going to use the magic word outside of the template? It is of course technically possible, but what purpose would it serve, and how likely is it to happen? As a logical extension of that thought, what would the consequences be of using the magic word more than once in an article? DannyH (WMF), can you elucidate? Fram, I have tested the CSS in Windows 10 for Chrome and Firefox, probably latest versions, it seems to work for the usual skins, and it is so simple that I can't see that it is likely to have any side effects, but everyone should feel free to test it in other operating systems and browsers just in case. I will leave it a day or two in case there is further comment, then post on Village pump (technical). If anyone has useful input on a preferable display formatting, please speak out. It should be immediately obvious that it is not ordinary content, a tag, or a hatnote. Italics on light yellow background does it for me.· · · Peter (Southwood) ^(talk): 12:08, 30 March 2018 (UTC)[reply]

RexxS, The magic word now exists. When you have the time could you see if can be used from the template?

@Pbsouthwood: Sure, I'll rewrite the template when the magic word is working to create a wrapper initially for the magic word with a display class. That will allow us to see the short description by setting our personal css while we work on the full implementation. Unfortunately, right now, DannyH (WMF), the magic word does precisely nothing - i.e. the Wikipedia app still displays Guy Madison as "actor" (per Guy Madison (Q1557542)), not "American film and television actor" as in {{SHORTDESC:American film and television actor}} our Wikipedia article. The same goes for Wagon Train on the app, which shows "Television program" from Wikidata Wagon Train (Q772909), not the {{SHORTDESC:Western television series aired 1957-1965}} that Danny added to the article. There's no sign of the short description showing up in any search that I'm aware of either. It does appear when editing using the Visual Editor, although I haven't yet figured out how to actually make changes to the text using VE. Is that by design? --RexxS (talk) 17:02, 29 March 2018 (UTC)[reply]

RexxS: Yes, that's what I was talking about above. Now that the API works, the teams need to make the interface changes to show the new enwiki-hosted description in the displays: the mobile apps, Visual Editor and ContentTranslation. The work is being documented on this ticket: phab:T184000. Sorry this is taking a while, I'm trying to move it along. -- DannyH (WMF) (talk) 21:52, 29 March 2018 (UTC)[reply]

@Pbsouthwood and DannyH (WMF): Update: I've amended Template:Short description to call the magic word, so that the API call now returns local values wherever the template is being used: Special:WhatLinksHere/Template:Short description - that's 4688 transclusions to date. See [4], [5], [6], for example. I have enabled my personal css to display the template contents (see User:RexxS/common.css) so I can see what's in the template while we're testing. Let me know if any problems arise. --RexxS (talk) 18:08, 29 March 2018 (UTC)[reply]

@RexxS: Is there a way to access the local values via Module:WikidataIB's getDescription so that it can be used in e.g., list articles, on Commons, or elsewhere? Thanks. Mike Peel (talk) 18:23, 29 March 2018 (UTC)[reply]

@Mike Peel: See Template:Short description/test. I'm offended that you thought I hadn't already implemented it It makes use of getDescription() from Module:WikidataIB. I can tailor it more specifically if you give me an exact spec or precise usage-case. --RexxS (talk) 18:38, 29 March 2018 (UTC)[reply]

Bah, I didn't read carefully enough. You mean access the short descriptions we write in English Wikipedia in other projects? That would be interesting, but would require an extension to Extension:Scribunto to get values via the API. Phabricator ticket and a lot of arguments, I guess. --RexxS (talk) 18:45, 29 March 2018 (UTC)[reply]

What about within English Wikipedia? Can we call up a short description for use as an annotation in a list or category? · · · Peter (Southwood) ^(talk): 19:05, 29 March 2018 (UTC)[reply]

It's the same issue - we need to get the data (short description string) back from wherever it's held on the server's database. Generally, it's not practical to try to retrieve any old piece of text, but this is a special exception that the API recognises as prop=description. It's really intended to be accessed in PHP, or as a stand-alone call in an external program, rather than something embedded in a Wiki page. I might be able to use AJAX inside JavaScript to fetch it within a Wikipedia page, but I couldn't then push the JavaScript to every reader. It would really need an implementation in the MediaWiki software.

Thinking laterally, one possible solution would be to create a (preferably read-only) property on Wikidata that imported the "short description" from each language Wikipedia as monolingual text. That could then have qualifiers, references, etc. and we could access it using the usual tools. Another controversial solution would be to insist that where a Wikipedia article with a short description exists in a given language, the Wikidata description in that language has to be taken from the Wikipedia short description. A bot could do that. Just thinking aloud, of course. --RexxS (talk) 22:30, 29 March 2018 (UTC)[reply]

If I understand you correctly, that would be using Wikidata to store a copy of the short description in a form which could not be edited on Wikidata, but would be a true representation of the item on Wikipedia, because it is easy to access from Wikidata. That could be workable if the Wikidata people are amenable, and the update lag at Wikidata does not make vandalising the short description at Wikipedia attractive. It would be better if it could be done more directly, but getting an implementation in Mediawiki might take forever. A short description is potentially very useful for a variety of applications, but only if it is accessible. It looks like what I want to do and Mike Peel wants to do are effectively the same thing. Bluerasberry was also asking about short descriptions for use in translated material for other projects. I don't know if this has any bearing on that application. The Transhumanist might also be interested - Outline and index articles could be an application for using short descriptions as annotations called from within another article. They would be heavy usage, as several hundred calls could be made from the one article, but cacheing might manage that problem. I don't know. Things that are not possible now may become possible sooner or later. With more interest and more applications it may be sooner. · · · Peter (Southwood) ^(talk): 06:26, 30 March 2018 (UTC)[reply]

• Documentation request to WMF staff Danny said, "I think that RfC would be most likely to succeed if it came from editors, rather than the WMF. Can I do anything to help?" Yes! I request that the WMF dedicate more staff time to documentation. (Actually - I encourage the WMF to be more liberal in offering grants for wiki groups internationally to write more documentation, but accomplish documentation however you can.) When the WMF is in a hurry then it cannot expect for volunteers to match its schedule. There are lots of volunteers who are willing to participate in a discussion which is easy to understand, but when the barriers to understanding are high, and when there is not an obvious way for the community to have discussions, then that precludes the community enjoying a conversation on a topic.

The current documentation at phab:T184000 is not written for humans and is not a workable basis for general community conversation. Please simplify!

1. Establish page on meta. Name the project. There is meta:Help:Magic words. What is this project? "Magic word description from Wikidata on English Wikipedia"? Call it anything.
2. Link to all previous conversation and documentation
3. Link to the other related tools and documentation which puts this project in context
4. Define the project
5. Please give a go at listing potential drawbacks, or state that you have not identified any

If you can do this much, then when you give community notice, then there is a place to ask questions, confirm support, voice concerns, etc. Right now none of the usual conversation is possible.

I recognize that usually the wiki community establishes conversation but it seems like there is a time schedule to meet here. Having someone who understands the project start the on-wiki documentation is very helpful. Typical wiki community members cannot post to phabricator.

I can only recruit conversation participants on topics which are possible to understand and discuss! This currently is not one of those! Blue Rasberry (talk) 15:08, 30 March 2018 (UTC)[reply]

So, we have the WMF position ("first populate 2 million + articles with good descriptions") and the RfC result ("start with blanks, blanks overrule Wikidata descriptions") which are not compatible. The effect of the WMF dictate seems to be that no one is interested or motivated to start the actual process of adding the magicword (if an when it arrives), never mind populating it. Which would mean that two RfCs and the development of the magic word are largely wasted, apart from individual, slow, local additions of the magic word (which are welcome, don't get me wrong).

I was thinking of starting a bot request to get this going, but the addition of empty magicwords seems a bit pointless if we will have to let a bot run across half the articles again to populate them anyway. So what we need, unless people want to force the WMF somehow to implement the result of the RfC, is good, bot-applicable descriptions. The easiest way would be to simply copy the Wikidata descriptions (excluding things like lists, disambigs, ...), but of course this would import all problematic ones as well. So all better ideas, preferably feasible in a relatively short time, are welcome! Fram (talk) 11:21, 27 March 2018 (UTC)[reply]

Make it visible in the desktop version using a gadget and asking people to set it seems the most logical approach to me. —TheDJ (talk • contribs) 14:33, 27 March 2018 (UTC)[reply]

That is not an offer by me to write it btw.. —TheDJ (talk • contribs) 14:35, 27 March 2018 (UTC)[reply]

Thanks, it is something that is probably needed, but doesn't seem the way to populate 2 million of them... Fram (talk) 15:01, 27 March 2018 (UTC)[reply]

I might have missed something, but how many/which ratio of the currently existing short definitions on Wikidata are bad? Jo-Jo Eumerus (talk, contributions) 15:21, 27 March 2018 (UTC)[reply]

@Jo-Jo Eumerus: Check the archives of the 2017 page, eg. here. May possibly have been discussed during the VP debates as well. Nikkimaria (talk) 01:00, 28 March 2018 (UTC)[reply]

First I think we need to be clear on the current situation. The original RFC was overwhelming to remove the Wikidata descriptions, and the WMF agreed to turn the wikidata descriptions feature off for enwiki.[7] The WMF then failed to do so. A second RFC to remove Description field from Wikidata in any view of en-WP was then withdrawn, based largely on my own vain hopes that the WMF would be "willing to clean it up amicably". We then spent months on negotiations, trying to to give the WMF what it wanted... trying to find a way to 'fix' descriptions instead of removing them. This let to the suggestion that we might take on the work of writing local descriptions. A third RFC simply asserted that result by implication, and asked how to implement it. The WMF then threw the entire RFC result and all the months of negotiations in the garbage. So right now, we have squat except the original original consensus to eliminate wikidata descriptions. I see no reason anyone couldn't or shouldn't re-open the second RFC. Wikidata descriptions are already removed from mobile-web, and it's unclear why mobile-app should be any different. These search results use the lead text of the article, and it's unclear why we would need different and contradictory search results.

If we go forwards with a bot run and writing millions of descriptions:

• The first question is whether to limit it to just all article pages? The WMF is still insisting on shoving out Wikidata by default. Wikidata has items linked to a lot of non-article pages, and Wikidata can create descriptions for any page at any moment. Given the non-cooperation of the WMF, the only way to fully remove Wikidata descriptions would be to add the magicword to every page everywhere. Or are we only concerned with articlespace?
• I think the best way to get descriptions filled is to having a preview-only message asking for the description to be filled in. The message would only appear if the description is empty. The deleted template {{reflistp}} had similar preview-only functionality, and I was planning on porting it to the new {{short description}}.
• The bot could roll out a default value for the magicword, but I think this is better handled via the template. It should be simple enough to avoid the WMF's content-filter and have the template shut off wikidata values by default.

Alsee (talk) 22:13, 28 March 2018 (UTC)[reply]

Alsee, The first things we need are a way of seeing what WMF are actually using, and what the current magic word actually is. Cheers, · · · Peter (Southwood) ^(talk): 06:51, 29 March 2018 (UTC)[reply]

Are there any known or foreseen problems with Wikidata descriptions of non-article pages? If so, then a simple generic type description such as "Wikipedia article talk page", "Wikipedia user page", "Wikipedia template page" etc should suffice. After all, they are not part of the encyclopaedia and we have no special interest in making them more accessible to non-editors, and one would assume that WMF would not need a more detailed description for any legitimate purpose. A bot run can do this quite easily.

Your suggestion for a preview only message seems like a possible way forward. It would be better if it can also be displayed by an opt-in gadget in normal reading view, and this already exists for the short description template. If the magic word can be made to work from inside the template then we may have this part of the problem under control.

I agree that the template is a better way of managing the magic word, if it can be managed by template, which has yet to be demonstrated. · · · Peter (Southwood) ^(talk): 07:21, 29 March 2018 (UTC)[reply]

The disambiguation page crew would like the magic word to be deployed by embedding it in the disambiguation templates, which seems like a good idea and might be a good place to start. · · · Peter (Southwood) ^(talk): 07:28, 29 March 2018 (UTC)[reply]

A proposal for embedding the short description in the disambiguation template is proceeding at Template talk:Disambiguation, as the first tests seem to work. · · · Peter (Southwood) ^(talk): 09:37, 31 March 2018 (UTC)[reply]

Couldn't we do the whole thing locally? Have an index file with article titles paired with their descriptions? And have the magic word initiate data retrieval from that? Both Wikidata and the WMF sound like they are near impossible to work with. Leapfrog them. Wikipedia has the talent available.     — The Transhumanist    11:10, 30 March 2018 (UTC)[reply]

I think that ship has already sailed, but yes, cooperation with WMF and Wikidata is complicated. Our aims are not perfectly aligned. · · · Peter (Southwood) ^(talk): 09:37, 31 March 2018 (UTC)[reply]

Hi all, here's an update on the display of short descriptions. The link dialog in VisualEditor and New Wikitext Mode is pulling from the local description override now. (If you want an example, Guy Madison has "Actor" as the description in Wikidata and "American film and television actor" as the English WP override.) It looks like the overrides will also show in the iOS and Android apps within the next couple of weeks -- if anyone's interested, the iOS ticket is T191347, and the current Android ticket is T191337. So things are moving forward, as we work through the different instances of where the description is used. Let me know if you have any questions or thoughts. -- DannyH (WMF) (talk) 20:53, 30 April 2018 (UTC)[reply]

And another update: the short descriptions override is now appearing on the iOS app. The Android app should be updating very soon, and I think that's the last big piece. -- DannyH (WMF) (talk) 17:35, 14 May 2018 (UTC)[reply]

Considering this is for ~500000 descriptions, people may be interested in Template_talk:Infobox_settlement#short_descriptions Galobtter (pingó mió) 16:06, 14 April 2018 (UTC)[reply]

Template talk:Infobox album could also add 140,000 pages. I started a discussion there, but haven't produced any code yet. Other potential candidates may be Template:Infobox song (60K), Template:Infobox book (40K) Template:Infobox film (120K) Template:Infobox musical artist (100K) Template:Infobox radio station (20K) Template:Infobox television (40K) Template:Infobox video game (22K) Template:Infobox NRHP (63K), ... Fram (talk) 15:04, 25 May 2018 (UTC)[reply]

Hey, just as a minor related update: It’ll be possible to edit already existing local English Wikipedia short descriptions in the version of the iOS app (6.2) planned to be released in January. This is currently not the case. /Johan (WMF) (talk) 08:42, 11 December 2018 (UTC)[reply]

See phab:T206786 for more information. /Johan (WMF) (talk) 08:43, 11 December 2018 (UTC)[reply]

This is now included in the latest beta release of the iOS app. /Johan (WMF) (talk) 07:52, 14 January 2019 (UTC)[reply]

(Lacking a 2019 page.)

Hey, just as a small related update: It’ll be possible to edit Wikidata descriptions in the Wikipedia Android app. The local descriptions will always be prioritised and there will be no link to edit the English Wikidata description for articles that have local descriptions, to avoid confusion. This is related to the app editor tasks project that has previously been reported in Tech News. /Johan (WMF) (talk) 18:04, 23 April 2019 (UTC)[reply]

The new version of the app is currently scheduled to be realsed towards the end of the week. /Johan (WMF) (talk) 18:06, 23 April 2019 (UTC)[reply]

Also, if you think this should definitely be posted somewhere else, please tell me. I don't think should affect much beyond the app editor tasks (and doesn't affect the mobile web), which has been developed for the Wikipedias in general, but the app has taken the English Wikipedia consensus into consideration to make sure we the agreement to prioritise local short descriptions is respected. Should you encounter any issues (which you shouldn't, but you know, just in case), you can report them on mw:Talk:Wikimedia Apps/Team/Android/AppEditorTasks (or ping me). /Johan (WMF) (talk) 18:20, 23 April 2019 (UTC)[reply]

This is now active. /Johan (WMF) (talk) 16:02, 29 April 2019 (UTC)[reply]