I think it is not about a file being or not the best to reprensent something, it is about "can commons gain something by being in wikidata". And I think it is the case : wikidata will give very powerful tools, with properties to classify and describe files, and queries to find which files matches some criteria which could help users to find what they want.

Of course the amount of work to describe precisely each file is gigantic, so this system will not be available at his full potential, but we can be sure that we can build something stricty better that the current category system with a very limited cost since wikidata is already there and is planned to interract with commons. And the availability of wikidata items will for example a very fine graine matching of some schéma with their topic, so we can for example find every illustration used for a mathematical topic, whatever it is about a very specific theorem, work can be tedious to do with the current system as the classification is higher grained.

2013/6/22 Jane Darnell <jane023@gmail.com>
I was thinking about items vs properties and Commons. I am not sure a
"F" entity is necessary. In theory, each file on Commons can be linked
to another one, and each item on WikiData can be linked to another
one, but those links do not necessarily need to interconnect with
Commons. If a Commons file is not the "best promoted image" of a
certain WikiData item, then in my mind it does not need to be in

Here is an example of what I mean:
This is an image of an engraving of Dirk van Hoogstraten:
He has a WikiData (person) item that could link to that image here:
The image file is derived from this file:
which is a page from a 3-volume book about artists that has a WikiData
(book) item (that could use the titlepage of the first volume as
linked image). The picture of Hoogstraten on that page was also used
later by another engraver in a later book about artists:

This second book also has a Wikidata (book) item, and the authors of
both books have WikiData (person) items, and so do the engravers of
the original engravings. If the original drawing of Dirk Hoogstraten
ever comes to light, then that image could be promoted to "best image
of Dirk Hoogstraten", and this one can remain on commons, but does not
need to be linked to WikiData, or do you think there is a need for
this file to have its own "F" status in WikiData?

2013/6/21, David Cuenca <dacuetu@gmail.com>:
> On Wed, Jun 19, 2013 at 11:15 AM, Denny Vrandečić <
> denny.vrandecic@wikimedia.de> wrote:
>> I agree that the different projects have different requirements. But I
>> think we should strive for a small number of "Wikidatas" - you could have
>> made the same argument for Commons, after all.
> The two projects that might need it most are Wikivoyage (hotel/restaurant
> listings) and Wikiquote (semantic quotes), though in the end they could be
> included in Wikidata with different entity types.
> Wikisource is aligned with the existence of source items in Wikidata, so
> other than having a "S" namespace for sources (or not) and adapting the
> extension to work in Wikisource, there is not much need of development from
> the Wikidata side. See more here:
> https://www.wikidata.org/wiki/Wikidata:Wikisource
>> Right now, I think there is a need for Commons to have better support for
>> data - we are working on a proposal text for that - and Wiktionary - as
>> it
>> is really a different system - needs some special treatments - we have
>> just
>> send a proposal text for that.
> The Wiktionary proposal is a great start. I'm excited about the Commons
> proposal. It would be fabulous to have an "F" entity type for files :)
> (You can always go further and say "but it would be better if we supported
>> Wikibooks with structured data about the books and its chapters" etc.,
>> but
>> at some point you need to weigh implementation effort and cost and the
>> expected benefit)
> Wikibooks (user-generated text-books) is a special case, a bit different
> from Wikisource (digital versions of existing sources).
> However both can be treated in a similar way. I agree that the potential
> benefit of linking chapters wouldn't be that high.
> Cheers,
> Micru

Wikidata-l mailing list