I don’t fully understand this either; Wikidata is pretty complicated, and there’s quite the learning curve. However, I strongly suspect that the reason why can be summarized much the same way many Musicbrainz limitations are: “Currently the backend doesn’t support that, but we’re working on having it in a future version”
Is this just a case of “Hey! Person A in your group believes Statement X, but person B in your group believes the contradictory Statement Y. Why doesn’t everyone in your group believe the same thing?!”
Is there a Wikidata bug tracker where we could submit the issue, or follow it up, if it is already reported?
I see that having this long list of Wikipedia links is very awkward. I agree that it is kind of a workaround, and I would strictly only apply it in cases where the Wikidata item exists but does not provide that link. Wikipedia links can be very useful when doing further research on an item. Ultimately this should be fixed on the Wikidata side (e.g. by allowing multiple items to link to the same page). I think that’s something we can all agree on.
Back when we still used only direct Wikipedia links we used to only link the English version (if it existed) and a specific language version that was most fitting for the item in question, usually matching the artist’s country or release country. E.g. if it’s a Swedish artist the Swedish Wikipedia entries might be the most complete ones. Maybe that would be a way to make this manageable.
I wouldn’t say that it was manageable.
I often had broken links doing this way:
I think I’ve been perfectly clear. But why do I have to pick a side?
WD data is often wrong - at least, in terms of how it fits into the MB structure. In fact, it is often wrong as far as how WD wants it, too.
MB should not corrupt MB’s data simply because someone wants to use corrupted WD data.
So which one do you mean?
- We should not link to WD
- We should link to any WD page that provides interesting WP content for this MB entity
- We should only link to WD of the same type as MB entity, even if it means getting no interesting WP content
This is the only true option - linking to correct data.
If the data is not correct, why would we ever consider linking to it. All that does is weaken our data.
I don’t completely agree.
This would be the ideal situation, but what I want is (in order of importance but all together if possible):
- The link is not too prone to be broken (this excludes Wikipedia, that frequently renames or deletes its articles) *
- MB work page shows interesting relevant article about the song **
- MB release group page shows interesting relevant article about the single **
The way to get what is important to MB, is to link the only Wikidata that is linked to our interesting article, because we have no hint that they will ever solve their problem of not being able to link 2 differently typed WD to 1 combined WP article.
* I was fed up, long time ago, of having to change WP links each time I realised they broke. It was a pure loss of time and energy.
** As soon as it is created (in the future). As long as it persists. As soon as it is created again if it had been removed for being non-notable, for example. Even if it is renamed, frequently happens to the text between brackets. This also excludes Wikipedia (same reason, actually )
Heh… part of the learning curve I mentioned, is (above and beyond learning things) is figuring out where to go for information when you want to learn a specific thing about WikiData, and figuring out whether/where to ask a question or if the answer is already documented somewhere. This is where I often run out of motivation.
That said, I think this issue breaks down into two pieces (which might be in two different bug trackers):
- Get the functionality added to Wikibase (maybe you can find a info about a bugtracker there)
- Get WikiData updated to the new version of Wikibase that has the functionality.
The Wikimedia issue tracker is found here: https://phabricator.wikimedia.org
A relevant ticket might be ⚓ T152465 Investigate options for "one site link per item limitation" problem
T152465 is backwards to the issue. It is Wikipedia who needs multiple Wikidata links. Wikipedia has both Song and Single on one combined page whereas Wikidata is splitting this.
This is really a MusicBrainz issue to solve. We can’t tell other databases what to do. We need to find some way to still enable MB to show the rich human readable info of a combined Wikipedia page when the Wikidata link does not hold that link.
Wow, a nikki ticket!
Indeed it seems to be about the reverse, several WP to one WD.
But we need several WD to one WP.
Yes indeed and we cannot wait. We must link.
So we can just go on doing like we have done since the beginning of the RG to WD and Work to WD links.
Allowing WD Singles and WD Songs being linked to MB Songs (Works) and MB Singles (RG).
We always have such mappings to do when we want a new external DB in the whitelist. Because why would every external database be miraculously on the same model as MB?
Trying to find an actionable immediate solution that doesn’t need constant monitoring of this linked URL and that didn’t need to bend concepts too much.
So we know how Wikipedia pages are.
They regroup work, release and recording information.
They also include covers of the work (other artists who are playing or singing the same work)
They are interesting like that.
They will rarely be split (not seen yet).
So what it is in fact?
It’s all about a work and its derivatives, isn’t it?
And, as a Wikipedia can only have one Wikidata (type), let it be a work typed Wikidata:
- We link MB Work to WD Work
- If this WD Work is an A side of the single, then we can link the MB Single Release Group to this WD Work
- We never link WD Single to any MB Entity
- If such a Wikipedia is linked to another type than Work Wikidata, we change this and make the WD Work/Song to WP page link
It means editing WP and/or WD pages (I don’t remember which)
So the supported types of Wikidata would become:
- Group of persons
And never Single.
Please, do not go change Wikidata to make it better fit MusicBrainz.
That seems like a good way to make the life of people trying to improve music in Wikidata worse. By doing that, you make it so that bots will probably repeatedly add the RG link to the WD work page, while the WD single is left empty.
It’d be nice, but we don’t need to do so - and any alternative that involves making wrong links to Wikidata, making the actual data worse just so we can display a Wikipedia fluff on single RG pages, is something that is not going to be style-endorsed.
Sorry, my misuse of words. Not “need”, maybe a “it would nice to see”. Not demanding anything. Just saying that we should not be breaking someone else’s database because it doesn’t fit the way MB can use the data.
By such a Wikipedia, I mean an article that documents a work and its recordings, releases (as Single or other), covers, versions, etc. The typical Song Wikipedia article.
What is wrong in typing such a Wikipedia as Work (or is it Song?) rather than the limitative Single type?
Isn’t it a legitimate Wikidata edit in its full rights?
It looks like a dead end, then.
Nothing, unless it goes against the wishes of people organizing music in Wikidata. If that’s how they do it, then that’s fine, just respect their style decisions.
I found this:
If a Wikipedia article is about a single with only one track, or mainly discusses one track of that single, then it should be connected to the item about the composition.
Maybe I’m not understanding properly, but if it’s a big deal for MB then we could easily tackle this in the display:
Show Wikipedia pages for songs on RG’s with type ‘single’ that contain that song. Maybe with a note, or more hidden than usual (to set it apart from a ‘true’ Wikidata link for that single). It would have to allow for multiple songs potentially being on the single.
Maybe collapse the whole WIkipedia heading by default if it’s too much.
The main problem with that is that it’s likely to pull up a lot of pages about the song and other people’s related singles but which don’t reference the performing artist at all, or at best just a mention ("… also covered by this artist"). Eg, just as an example, Blue Suede Shoes. Most people know the Elvis Presley version, and some might even remember Carl Perkins (I’d forgotten). Musicbrainz actually has 585 recordings of it (many duplicates but dozens at least distinct artists) but the Wikipedia Song entry mentions just those two.