Wikipedia talk:Requests for comment/Wikidata Phase 2

Top
I hope I am doing this right, and making the introduction and the options a bit more clearer in my opinion. I am still not completely happy with how it is phrased, but I hope we can find consensus on this. I will be available on #wikipedia-en and a few other channels for the next six hours or so. Cheers --denny vrandečić (talk) 09:53, 22 April 2013 (UTC)
 * Listed on WP:CENT. MER-C 12:08, 22 April 2013 (UTC)
 * Hm, I would have preferred to let Risker take a look at my reformulations, before it goes wider. I hope I didn't change her intended meaning, but I cannot be sure. --denny vrandečić (talk) 12:16, 22 April 2013 (UTC)
 * I've just had a chance to look at it now during my lunch break, and will make a few changes, although I agree with most of Denny's changes. Unfortunately, working in the non-wiki world, I do not have access to IRC until I get home tonight, and I assume that Denny being a sensible person will long be offline. :-)
 * Denny, I think some clarity is needed in a couple of places. I've not found any documentation or even suggestion from anyone I've spoken to that a change made on English Wikipedia would be reflected on Wikidata, yet some of your recent edits suggest this is the case. I've also found no documentation for your comment that changes to a Wikidata-generated field made on Wikidata would result in a watchlist change on English Wikipedia.  Now, that may be simply because I'm not looking in the right place (the documentation is spread all over the place and much of it is written in technicalese so is difficult for a non-tech to parse), but that doesn't seem to correspond with even the information that Wikidatans understand.  Some links to this information would be useful. They can be here on the talk page, or added to the "see also" section.  Risker (talk) 16:12, 22 April 2013 (UTC)

It really depends on what you mean with a "a change made on English Wikipedia" - what I mean is that we are eventually planning for UI widgets that allow you to change Wikidata from within Wikipedia, and where you need not to actually go over to Wikidata in order to do that. But this is indeed in a vague phase yet, and how these widgets would work heavily depend on the actual workflows of both communities. But we already provide the API that allows that and I would hope and expect that some volunteer JavaScript wizard will create such widgets - and actually, the first such widget is on the way and developed mostly by a volunteer contributor.

Regarding the changes to be displayed on the recent edits, that is described on Meta. But I admit that it is quite technical. In short: yes, your watchlist get notified when a change on Wikidata is relevant for the Wikipedia article you watch.

I hope this provides a bit the clarification you are seeking. --denny vrandečić (talk) 16:33, 22 April 2013 (UTC)
 * Well, it helps. First off, this RFC is about Wikidata in its current formulation, not what it hopes to be when it grows up. Thus, saying that it is *now* possible to make an edit on English Wikipedia through any UI that is reflected on Wikidata is not true; it's something that is hoped for and being designed for, but we aren't there yet.


 * As to the "watchlist changes" issue, no wonder I couldn't find it. :-) I understood about three sentences and a few occasional words. Bottom line:  Is this operational today, or is this another thing that is in the design phase?  Risker (talk) 16:42, 22 April 2013 (UTC)
 * yes, it is operational, I use it on my watchlist. There is a link above the watchlist you cn click, or change the preferences to have it on the permanent basis.--Ymblanter (talk) 16:45, 22 April 2013 (UTC)
 * Thanks, Ymblanter. Now I know what that link is for. Risker (talk) 16:54, 22 April 2013 (UTC)

Option 4
We need to clarify this. My understanding was that there was a process that called Wikidata to the template *only if* there was no current content in the parameter, and that is what I am after in this option. It is retrogressive to have Wikidata override English Wikipedia data and then require editors to go back and put the "old" English Wikipedia value back in. If I have misunderstood, and there is no opportunity to "keep" Enwp data in fields rather than having them overwritten and then have to change them back, please advise promptly, and Option 4 will be removed, since it isn't actually an option. (Incidentally, it's the most popular one so far.) Risker (talk) 17:41, 22 April 2013 (UTC)


 * Templates can and should be written in such a way that the locally given template parameter is used, overriding anything from Wikidata. Only in the case there is no locally given template parameter, Wikidata would be asked and the value from Wikidata displayed. --denny vrandečić (talk) 17:47, 22 April 2013 (UTC)
 * And that seems to the intent of option 4. However, I'll note that my suggested option "4.5" should also be a plausible option. Some fields just don't need overriding whatsoever. --Izno (talk) 17:58, 22 April 2013 (UTC)

Risker, your "tiny" change created the sentence "Modification to any Wikidata content in the field could be done either centrally at Wikidata, or by adding information to that field locally, which would override the data from Wikidata." I am not sure if I understand, but I won't to explain that Wikidata content is not being changed by adding a local value to a template in Wikipedia. Adding a local value merely overrides the value from Wikidata for display in Wikipedia. May I remove "any Wikidata" at the beginning of the sentence? --denny vrandečić (talk) 18:03, 22 April 2013 (UTC)
 * I hear what you're saying, Denny. How about "Fields using Wikidata can be updated centrally at Wikidata, or can be replaced by local content." Does this work for you?  Risker (talk) 18:07, 22 April 2013 (UTC)
 * Yes, that would be correct. --denny vrandečić (talk) 18:08, 22 April 2013 (UTC)

Use of the word "ban"
Sorry, Denny, but I have reverted to the original. On this project, an editor doing something that is "banned" (whether generally or just for that individual editor) is usually blocked. I don't think we want to escalate a prohibition or a community consensus not to do something into a blockable offense unnecessarily. I'm of the opinion that unnecessarily adding emotional language to this RFC is unhelpful (including the "even when a Wikiproject or editing community wants to"). Status quo should not be treated as deprivation. Risker (talk) 17:52, 22 April 2013 (UTC)


 * I have repeatedly tried to make clear that Option 1 means a global ban on any usage of Wikidata Infobox data in Wikidata infoboxes (or text, respectively). The comments by Risker say that I am "slanting" it with my edits. I want to see it explicated that Option 1 is a global rule for Wikipedia, and that no Wikiproject or other editing community could override this without breaking this RFC in case Option 1 is chosen. I think it is only fair for editors to make this implication explicit. --denny vrandečić (talk) 17:58, 22 April 2013 (UTC)


 * I moved the explication to the discussion. I hope this is acceptable, also in the way it is phrased. The discussion on Option 1 was so short anyway, compared to the others. --denny vrandečić (talk) 18:07, 22 April 2013 (UTC)
 * I still think you're making out as if a community's decision is going to actively deprive groups of editors from a precious jewel, but then I guess you've got to believe in your product the project you are steering. :-) However, I strongly urge you to remove the "Commons" bit: again, we use a huge number of local images on this project (especially in infoboxes) because we permit non-free use in limited circumstances. Many editors on this project have learned the hard way that remote content means lack of control of that content: I personally have seen hundreds of examples of images that got overwritten or deleted or otherwise badly modified at Commons. It's not as important when it's the sixth image on a long page, but when it's the infobox image it's rather a big deal. Please keep this in mind during the discussion: English Wikipedia doesn't feel the same way about Commons that a lot of other projects do.  Risker (talk) 18:19, 22 April 2013 (UTC)


 * I am unhappy with Wikidata being called "my product" - I am not here to sell it to anyone, and it is not developed as a product, but as an enabler for sharing knowledge. If I wanted to sell a product, I would have settled for Semantic MediaWiki long time ago.
 * Regarding Commons: we are learning also from the errors of Commons, and as said, we are displaying changes in your watchlist, etc. But still: one of the fundamental principles of the Wikimedia projects is that more eyeballs remove more errors. So, a vandalism on Commons - or Wikidata, for that matter - will be discovered and dealt with quicker than vandalism on any, especially the smaller projects. It would be a pity if the English Wikipedia choose not to join forces and share their most important resource - editor attention. --denny vrandečić (talk) 18:54, 22 April 2013 (UTC)
 * See now, there's a good example of the variable use of a term. I have lost count of the number of times members of the WMF engineering department refer to the projects they work on as their "products". Doesn't mean they're selling them, it means they're producing them.  Nonetheless, I have stricken the term and have replaced it with "the project you are steering", which I think is probably fair?  Speaking personally, I don't see Wikidata as inherently bad or wrong, and I can see some parameters where it could be useful. I'd like to see editors testing it here in a way that doesn't directly affect article space, at least in the beginning while Wikidata learns how to manage important parameters that involve numbers; right now it's so immature that it's more likely to be crushed by the Enwp behemoth than nurtured.  As I've mentioned before, nobody really cares all that much about language links, which were largely managed by bots for many years, so that was a brilliant first phase. However, editors here care a great deal about infoboxes and templates so finding a way to expose them to this new technology in a way that intrigues them rather than annoying them or inciting another round of edit warring is a net positive; with all the Enwp volunteers working on Wikidata, I am sure that the project could be shown in a positive light.  Perhaps a good first step, which can be taken immediately and without any kind of RFC, would be adding  templates to articles. I know that "deploy early, deploy often" is a watchword for IT, but in reality it is a big turn-off for end users who haven't voluntarily agreed to be test subjects, especially when it comes to changing work to which people have devoted a lot of time and effort. Risker (talk) 19:23, 22 April 2013 (UTC)


 * English Wikipedians are not test subjects. It has been switched on on other Wikipedias before. But in order to grow together, in order to see how Wikidata will be used, we should refrain from a project-wide prohibition of its usage. Basically, the fear you express is: Wikipedians are not mature enough to responsibly use this new feature, so let's tell them they cannot. Yes, it is not completely done yet, and major features are still missing -- but there are use cases already where the English Wikipedia could benefit from using Wikidata data, e.g. removing the number from the IMDB template, or shortening a few template calls. Switching on Wikidata does not mean that everyone should mindlessly use it as much as possible everywhere, but I'd say it would be a shame if the editing communities were not allowed to use it at all. --denny vrandečić (talk) 20:25, 22 April 2013 (UTC)
 * No, that's not what I am saying at all. What I am saying is that this new project is not mature enough to withstand the rigors of English Wikipedia, and because it is so immature, it will likely be poorly regarded and considered not useful. I personally am leaning toward Option 2, permitting testing on this project in article subpages, which has several advantages: it does test the software, which will identify bugs and discrepancies; it allows "regular" Wikipedians to see how it works without being worried about the quality of content in the articles (a non-threatening introduction to a new idea); its presence on this project is more likely to encourage other Wikipedians to participate in testing (much more likely to test if it is here than on another project) and possibly even adding data. I agree with several others that until Wikidata is able to manage the most common parameters (which it does not do now), it's not ready for article-space. Risker (talk) 20:55, 22 April 2013 (UTC)

Wording
I am afraid that if I, being an admin both here and on Wikidata, have difficulties understanding the options, other users, who are less involved with Wikidata, will have difficulties as well. Could we may be just take a template and a particular firld of this template, find the corresponding Wikidata item and explain what happens in all five options (including well-meant and vandal edits on Wikidata)? This should come as a supplementary info, not as a replacement of the list of options.--Ymblanter (talk) 18:05, 22 April 2013 (UTC)
 * Visuals would be a wonderful idea, Ymblanter. Perhaps links or screenshots from a test case on Test2 wiki? Risker (talk) 18:09, 22 April 2013 (UTC)
 * That would be great but goes beyond my abilities. I was thinking about smth more simple. I will be back here in a few minutes with an example.--Ymblanter (talk) 18:14, 22 April 2013 (UTC)

Basically, assume a template called Infobox city with the parameter mayor. If is displayed as an Infobox, giving the mayor Donald Duck. If the mayor parameter is empty, like this we would see the city Infobox, but, in most cases, without the line for a mayor.

Using Wikidata, the template can be extended in such a way that if there is no mayor locally given, i.e. if the template is called like this, then the template can ask Wikidata for the mayor, and if Wikidata knows, display it. Bonus: if the value gets changed on Wikidata, your infobox will update automatically. So if someone updates the data on Wikidata to Mickey Mouse, your infobox will automatically follow.

If you decide not to trust those pesky Wikidatians on this matter, you might locally give a value, just as you have done previously: This would still still show Donald Duck to be the mayor, no matter what Wikidata says. You can locally override the value.

What if you don't want a mayor to be shown at all? Also this is no problem. The template can be written in such a way that a specific value is understood to suppress this line completely, even if Wikidata would be able to provide the value, e.g. like this:

MediaWiki templates are very flexible, and all of these options - and quite some others - can be implemented. Wikidata was always developed as an offer: the Wikipedias can use them if they want to, they can decide for each template, for each infobox, for each article themselves whether they want to use data from Wikidata or not. --denny vrandečić (talk) 18:28, 22 April 2013 (UTC)


 * So, let me see if I have this straight. I'm editing away and I see an error in the infobox. If it happens today, I open the edit window, I look for the error, I correct it, and I save.  If the same information is supplied to the article by Wikidata, I will see...nothing. I won't even see the WD codes?  It will just look like a blank field?  Risker (talk) 18:35, 22 April 2013 (UTC)
 * But that is the same as with the interlanguage links. You see nothing in the page code, but can easily repair them.--Ymblanter (talk) 18:41, 22 April 2013 (UTC)
 * I don't think that is true. Editors can't edit what they cannot see. And we certainly don't want our infoboxes cluttered with every parameter having a little "edit pencil" beside it, as is the current status for language links. Risker (talk) 18:45, 22 April 2013 (UTC)

(ec) Let me try it like this. We have d:Q968 for Warburg, a city in Germany. The Wikidata item currently states that the head of local government (myor for brevity) is d:Q1466247 Michael Stickeln (no article on English Wikipedia). This is done by using the Wikidata property d:P6 "head of local government". At some point, the mayor will change, and then some (presumably, German) users will replace them on Wikidata. A vandal can also come and replace the mayor with d:Q34086 Justin Bieber. An undexperienced user can come and accidentaly erase the field, replacing d:Q1466247 with an empty entry. On the other hand, we have an article on Warburg; the article contains a customized infobox, which has a line |Bürgermeister = Michael Stickeln. Now, my understanding of the options is the following:
 * Option 1: English Wikipedia does not care what is on Wikidata. In particular, if the Wikidata entry is changed, this is not reflected in the infobox in any way. A user should notice that the mayor changed, and amend the infobox. At best, one can ask a bot to trace the Wikidata changes and report them on Talk:Warburg (not implemented yet).--Ymblanter (talk) 18:30, 22 April 2013 (UTC)

Funny that we came out with the same example. Great minds think alike.--Ymblanter (talk) 18:31, 22 April 2013 (UTC) Denny, I think your explanation is way better than mine, but we still neet ro relate it to the five options.--Ymblanter (talk) 18:33, 22 April 2013 (UTC)

I don't think Wikidata is solid enough to be trusted as the source of infobox content and frankly I don't even think this can be debated (for now!). But I see real potential in its ability to eventually notice discrepancies between information tidbits that should be identical on all wikis. Some of you may know a version of Q6823265, a database report based on the idea that an individual should either be dead on all wikis or alive on all wikis. It has been very successful in helping to keep articles updated (and in some cases in fixing vandalism quickly). With Wikidata, it's now possible to use this same idea to ensure that all wikis have the same mayor for a town (or other political office holder), the same date of birth for the mayor, the same province for the town and so on. A lot of people are worried that Wikidata will create problems that humans will be unable to fix and we need to take these concerns very seriously or Wikidata will be soundly rejected. But the model I describe showcases Wikidata as something that will report potential problems that humans can then decide to fix. Pichpich (talk) 20:19, 22 April 2013 (UTC)


 * For your information: test2:Template:Wikidata is supposed to do just that. For now on test2:Germany there is an infobox without parameters. If you fill one it will be shown, ignoring the Wikidata value. For the parameters you could use: see test2:Template:Infobox Country for instance nationalsong=Micky Mouse HenkvD (talk) 18:17, 23 April 2013 (UTC)

And so it begins...
This edit popped up on my watchlist. Is there any way to track which articles are using properties from Wikidata? A hidden category or Special page would be preferable.

Denny: these type of workflow issues were brought up here: Wikidata/Workflow. Has there been any progress on being able to track Wikidata properties per page and/or being able to track Wikidata property usage (at a binary level) per page? --MZMcBride (talk) 21:20, 22 April 2013 (UTC)
 * agreed a tracking category of some sort would be really good. -- phoebe / (talk to me) 21:51, 22 April 2013 (UTC)
 * I have left the editor a message asking him to come and participate in the discussion. I have no doubt he is acting in good faith; however, even after this brief period, the early consensus is not for direct editing into articles but instead management of templates. Risker (talk) 22:02, 22 April 2013 (UTC)
 * MZMcBride, I've linked 47288 which addresses being able to track these pages. Legoktm (talk) 10:07, 24 April 2013 (UTC)
 * I also just created Category:Templates using data from Wikidata that we can use as a tracking category to identify which templates are using data from Wikidata. Legoktm (talk) 10:12, 24 April 2013 (UTC)
 * There's no "early consensus" - we either have consensus or we don't - and I don't see any strong reasons not to be bold in editing: it's how the wiki works. Once a feature that allows centralised data to be included in articles is enabled, then you can bet that editors will boldly use it. I made two weeks ago so that I had an example to help explain to others what Wikidata Phase 2 does. It was also instructive in that it showed me that I had to create a wikilink manually, although I had assumed from the documentation that would be done automatically. It makes a single wikidata entry with multiple values much less useful and needs to be worked on (https://www.wikidata.org/wiki/Q905 Franz Kafka's "country of citizenship" was where I found that problem). Without the possibility to make test edits directly in infoboxes, it will become painfully difficult to modify templates that cope robustly with all data. If you want to tell me and other editors who are exploring the new capabilities to stop, you are going to have to come up with a more convincing argument than your feeling about "early consensus". An RfC is a request for comments, not votes, and I don't see any compelling rationales opposed to test editing in infoboxes at present. --RexxS (talk) 01:19, 9 May 2013 (UTC)

Generating references with Wikidata
Not sure if this is related to Wikidata Phase 2 or not, but reading through some of the documentation I am reminded of something I noticed with some references. At some point a system of referencing was developed where subpage templates contained the actual reference information (generated by a bot using DOI and similar identifiers), and this was transcluded to the articles in which the reference was used (an example is this edit to Walter Gardiner, with the citation information called from Template:Cite doi/10.1098.2Frsbm.1941.0046). This has advantages and disadvantages, but the one that annoyed me the most was the prospect of references with little 'edit' signs next to them (you can see one of these 'edit' links at Walter Gardiner). These edit links are a circular self-reference within the article that shouldn't be actually visible to readers (the main edit link is rightly outside the 'frame' of the article). (i) Am I right in thinking that one idea is that references will one day be generated using Wikidata, or is that a future phase? (ii) Will the default be to suppress the little "edit" signs next to all the Wikidata generated content, and to have the editability of these Wikidata fragments only becoming visible after the main "edit" button at the top of the screen is pressed? Carcharoth (talk) 00:44, 24 April 2013 (UTC)
 * This would only be for references referencing information in Wikidata. The referencing feature is still under construction, however, since there are some missing data types. --Rschen7754 03:09, 24 April 2013 (UTC)

Checking Wikidata-data against Wikipedia-data
For, I've proposed adding a parserfunction that would check whether the local data is equivalent to Wikidata and if not, adding a tracking category. I think this is something we can do for a lot of templates to help identify errors in both Wikipedia and Wikidata. Legoktm (talk) 10:09, 24 April 2013 (UTC)
 * James has enabled a more complex parserfunction proposed by multichill which is also being used on nlwiki. You can see the categories being populated at Category:Commons category Wikidata tracking categories. Legoktm (talk) 18:25, 25 April 2013 (UTC)

Defining "infobox"
Skimming the discussion, we seem to be drifting towards #4 (cautious infobox inclusion) and #1 (no article-text inclusion). As I noted early on, and a couple of others have picked up, this leaves a largish gray area around "informative templates".

By this, I mean things that aren't infoboxes - the big boxes in the top right of articles - but use templates to provide additional structure\information to the article; ones mentioned so far include commonscat, authority control and Attached KML. I feel that these generally fit alongside infoboxes in a conceptual sense - they're not prose and they're not navboxes - but the discussion has been pretty vague on this, and it would be good to have some idea of how this was originally envisaged! Andrew Gray (talk) 10:59, 26 April 2013 (UTC)


 * Is anyone here aware of m:Requests for comment/Interproject links interface, which is (partly, sort of) about transitioning interproject links to wikidata? Rd232 talk 16:51, 27 April 2013 (UTC)
 * Since I edited there several days ago, I am obviously aware of that RfC, but to my understanding it is about smth completely different.--Ymblanter (talk) 17:41, 27 April 2013 (UTC)

When will this end?
When is the poll going to end?Remember (talk) 00:38, 9 May 2013 (UTC)
 * See Requests for comment . --RexxS (talk) 00:53, 9 May 2013 (UTC)
 * Can I move that we start the process of ending this RFC? I don't think we are getting many more votes or comments on the issue. Remember (talk) 21:18, 13 May 2013 (UTC)
 * OK by me. HenkvD (talk) 18:14, 15 May 2013 (UTC)


 * ...and now closed, thanks to . Andrew Gray (talk) 19:25, 15 May 2013 (UTC)

Use Wikidata for Authority Control
I suggest use of WIKIDATA starting with simple things like Authority control. In many ways it is similar to Interwiki links which were replaced easily and most of Authority Control data is already imported in Wikidata through bots. What is your opinion? --Nizil (talk) 11:00, 13 May 2013 (UTC)
 * I've rewritten Authority control template to use wikidata as fallback. See this discussion. Tpt (talk) 08:12, 20 May 2013 (UTC)

Progress in implementation with Lua
Today, at the Lua on Wikimedia event at Wikimedia UK, I worked with RexxS and Justinc on using Lua to pull data from Wikidata. The code is at Module:Sandbox/Tom Morris. We created a modified version of Template:Infobox person at Template:Infobox person/Wikidata. It does one very simple thing: pulls the "spouse" parameter from Wikidata for articles. You can see it in action at Wikidata/Wikidata Sandbox.

I don't have any strong opinions about whether or not to switch over to using Wikidata for infobox data, but we have taken the first steps to showing how you'd go about doing it. It is done in such a way as you can override the showing of the parameter by either setting it locally or having an empty property. The code is rather ropey and hacky, but am happy to try and improve it before we start implementing it in infoboxes. This is something of a proof of concept: unlike using the #property template, you can use this to do complex analysis on combination properties (which I can explain if people are interested, but I now need a drink).

Hope this helps. —Tom Morris (talk) 15:31, 19 May 2013 (UTC)

So when can we start using wikidata
I see that the RFC has been closed, but when will we be able to use wikidata properties in infoboxes? Remember (talk) 16:35, 30 May 2013 (UTC)
 * No answer, I see. Well, this is not very helpful. Remember (talk) 12:54, 27 June 2013 (UTC)
 * You just need to start a discussion on the affected infobox page, and then do it if there's no objections. Template:Infobox road already does this for the map field; we're working on others, but it will take a while to recode as we're trying to get Lua and TemplateData in the same round. --Rschen7754 16:20, 27 June 2013 (UTC)

If this is policy it is out of date, where is the current allowable actions defined
(as the administrator whose closing statement is the current definition of the allowable action) ... Do you believe that this is still our policy point for use of Wikidata, and its use in infoboxes? If not, are you aware of a more up-to-date guidance on our policy per Wikidata (this has been quoted to me as defining where we are in December 2016 and that there is need for further discussion for further steps. It seems time to move to/consider phase 3. I am not sure that a static attachment to phase 2 is an appropriate reality. — billinghurst  sDrewth  06:55, 21 December 2016 (UTC)
 * Addendum, especially when one reviews that authority control is now parameter free, and we have removed "persondata". I will be damned if we need to walk through wikipedia and clear templates on a one-by-one basis. — billinghurst  sDrewth  06:58, 21 December 2016 (UTC)
 * Well, this was a solid consensus... well over three years ago.  I should think that how Wikidata has matured – and how integration has gone – warrants reexamining whether that very conservative approach remains relevant.  My personal take on it is that tighter integration and switching the defaults around would probably be a better approach nowadays; but there should be discussion since I've yet to be declared god-king of the project.  :-)  &mdash; Coren (talk) 16:20, 2 January 2017 (UTC)