To pick up on a few different comments from this thread:

@revi (Hong, Yongmin)

-- Yes, of course you are correct that it is categories rather than galleries that are the important structure for finding and navigating images on Commons.

But even if we were to change the status-quo and ban sitelinks from Wikidata items to Commons galleries (which might not be 100% popular, since we currently have 87,000 sitelinks to galleries, up by about 3000 in the last year) -- even if we were to ban sitelinks to galleries, this would still leave the question of whether Commons categories should be sitelinked to categories or to articles.

Currently there are 311,000 Commonscats sitelinked to category items on Wikidata, and 207,000 sitelinked to article items on Wikidata.  So the status-quo is for Commonscats to be sitelinked to category items (and in the past has been even more so).

The problem is that, the way the software exists at the moment, you can't have both.   So if a Commonscat is sitelinked to an article item, that precludes a Commonscat being sitelinked to a category item, and vice-versa.

At the moment, the expectation is that a Commonscat will be sitelinked to a category item, if possible.  Of 323,825 Commonscats that can be identified with a Wikidata category item, 311,000 are connected by corresponding sitelinks.  So if people are writing scripts or queries to look for such relationships, they will most likely look for sitelinks.

On the other hand, of 884,439 Commonscats that can be identified with article-like Wikidata items, only 207,494 (= 23.4%) are connected by sitelinks -- even if this number has doubled in the last 12 months, the expectation in the current status quo is that such a connection is more likely *not* to be represented in a sitelink, by a three-to-one margin.

Instead, links between Commonscats and article-like Wikidata items are currently overwhelmingly represented by the P373 property, which at the moment records 807,776 (= 91.3%) of such identifications.

This is the property that the script
   https://commons.wikimedia.org/wiki/User:Jheald/wdcat.js
looks up in order to display a link to Reasonator if there is a Wikidata article-item for a Commonscat.

To get the best idea of whether there is a corresponding Wikidata item and Wikipedia articles for a Commonscat, Commons users should therefore use   wdcat.js -- which is easily activated by adding a line to the common.js file, such as at
   https://commons.wikimedia.org/wiki/User:Jheald/common.js
-- because this will pick up four times more connections than currently exist as sitelinks.

The important think is to establish and preserve clear expectations, that software can build against.

At the moment, with the confusion of sitelinks of Commonscats to articles and to categories, there is no guarantee that it is possible to create a sitelink to an article.  On the other hand, it should always be possible to create a P373 connection.  They are the connections that systematically *can* be made, so it is important to make sure that they systematically *are* made, to drag up the return rate from the current 91.3% to nearer 100%.

That would be helped by a policy that was absolutely systematic in prescribing what should and what should not be sitelinked.


@ RomaineWiki:

You say that actively enforcing the longstanding Wikidata sitelink policy of only sitelinking Commons categories to category-like items, and Commons galleries to article-like items would be a plan to
"demolish the navigational structure of Commons".

But it wouldn't change any of the internal structures on Commons, and would merely underline the current fact that even now only 23% of Commonscats are linked to articles by sitelinks, compared to 91% by P373.

Isn't it better to get Commons users used to using (and improving) the wdcat.js  script, which uses the P373 property that can always be added, rather than perpetuating the current muddle of  Commonscat <-> article sitelinks, which are so haphazard ?


@ Steinsplitter

As I understand it, the long-term plans for a new Wikibase structure specifically for Commons are currently no longer an immediate development priority; but will presumably start to move forward again sooner or later.

On the other hand, this discussion was specifically about sitelinks.

Here I believe what has driven the Wikidata side has been the desire to have a rule that is simple and consistent and predictable, because that is the foundation needed to develop queries and scripts and tools and user-interfaces on top of.

Combining that desideratum with the technical restriction of only allowing one sitelink to each item from each wiki and vice-versa, is what has led to the recommended scheme of linking
  Commons categories <->  category-like items
  Commons galleries  <->  article-like items

with property P373 to handle identification of Commons categories <-> article-like items.

This fulfils the requirements of simplicity, consistency and predictability.

It's not ideal from a user-interface point of view (or a philosophical point of view).  But so long as the rule is applied consistently, the limitations it leads to can be worked round with appropriate software improvements -- eg in the first instance the   wdcat.js  script.

But to encourage people to develop and improve such software, it is helpful for the above structure to be applied consistently.

In contrast perpetuating inconsistency and muddle blurs what is needed, and works against the stable predictable basis needed to make such software work.


All best,
   James.









On 29/08/2015 14:39, Steinsplitter Wiki wrote:

Wikidata needs to ask the Commons Community before doing commons related changes.

It is so hard to understand what the wikidata people like to do with commons. Tons of text, hard to read.   I don't understand what they like to do, but if this change is affecting commons then commons community consensus is needed.

Date: Fri, 28 Aug 2015 17:34:28 +0200
From: romaine.wiki@gmail.com
To: wikidata@lists.wikimedia.org; commons-l@lists.wikimedia.org
Subject: Re: [Commons-l] [Wikidata] Trends in links from Wikidata items to      Commons

As I wrote before, that thought is too simple. You only say that a zero belongs to a zero, and a two belongs to a two, then you only describe the type of page, but you ignore the subject of a page. That subject matters much more than the namespace number.

Especially Wikinews is a wrong example, as most categories on Commons do not have a 1 to 1 relationship with Commons.
However, articles on Wikipedia do have mostly a 1 on 1 relationship with categories on Commons.

Romaine

2015-08-28 17:09 GMT+02:00 Luca Martinelli <martinelliluca@gmail.com>:
2015-08-28 12:09 GMT+02:00 Romaine Wiki <romaine.wiki@gmail.com>:

And I agree completely with what Revi says:

Wikidata ignores this Commons' fact by trying to enforce ridiculous rules

like this.

It's not such a ridiculous rule, if you think of the rationale behind

it: if gallery = ns0 and category = ns2, linking ns0 <--> ns2 in the

same item is IMHO not a rational thing to do (not even for Wikinews if

you ask me, but I'm digressing).



So the *practical* problem that we have to address is the list of

links in the left column. We really don't have any possibilty to

exploit P373 in any way, not even with a .js, to fix this?



L.



_______________________________________________

Wikidata mailing list

Wikidata@lists.wikimedia.org

https://lists.wikimedia.org/mailman/listinfo/wikidata




_______________________________________________
Commons-l mailing list
Commons-l@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/commons-l                                 



_______________________________________________
Commons-l mailing list
Commons-l@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/commons-l

_______________________________________________
Commons-l mailing list
Commons-l@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/commons-l