Hi Gregor,

thanks, this kind of input is exactly why I put these mails on the public list :)

2013/1/30 Gregor Hagedorn <g.m.hagedorn@gmail.com>

Some of our insights into the SMW RDF export (which we found to be difficult to configure and use):

1. Probably most relevant: total lack of support for xml:lang, which would have been essential to our purposes.

Wikidata should be planned with support for language in mind.

SMW does have support for xml:lang, but only for the one language the wiki is in. SMW was not intended for multilingual support.

In Wikidata we have a stronger focus on this, and will use xml:lang, but completely transparent to the user.

So, agreed.

2. We also found that we had serious problems with managing structure, e.g. record and subobject. Due to the need to obtain this information recursively by repeated calls, and because there is no control on the URI created for these calls, some easy solutions like applying clean-up xslt will not work. This may not be relevant for wikidata.

This should indeed not be relevant, agreed. (I wouldn't use XSLT here, by the way, but maybe a SPARQL construct query).

3. At first the lack of variable datatype (datatype is fixed per property) is acceptable. However, we found this a major problem with respect to the forced distinction between datatype:wiki-page and datatype:global URI properties. Essentially, SMW forces one to introduce for a semantic property (e.g. dc:creator) two distinct dummy properties: property:creator_page and property:creator_uri. Since in RDF export the artificial distinction between pages and URIs disappears, it would be desirable to merge them, but only one of them can be set to an imported vocabulary.

I think this may be relevant to wikidata, where a similar distinction between properties pointing to a local wikidata item and a global resource exists.

Hmm... this really depends on how we deal with URIs to external entities in Wikidata. To be honest, we have not decided that yet (or even thought much about it). The assumption was always more like "we have everything as an item", and the idea of external items has not really made it to the list yet.

Without that, this is not relevant for Wikidata. But especially for Wikibase installations that are not Wikidata this might be a very valid point. I have to think more through it.

Gregor

(PS: If any of the problems above in reality does not exist in SMW and we simply overlooked the solution, I am very happy for corrections, of course!)

--
---------------------------------
Dr. G. Hagedorn
+49-(0)30-8304 2220 (work)
+49-(0)30-831 5785 (private)
http://www.linkedin.com/in/gregorhagedorn
https://profiles.google.com/g.m.hagedorn/about

This communication, together with any attachments, is made entirely on my own behalf and in no way should be deemed to express official positions of my employer. It is intended only for the person(s) to whom it is addressed. Redistributing or publishing it without permission may be a violation of copyright or privacy rights.

_______________________________________________
Wikidata-l mailing list
Wikidata-l@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wikidata-l

--
Project director Wikidata
Wikimedia Deutschland e.V. | Obentrautstr. 72 | 10963 Berlin
Tel. +49-30-219 158 26-0 | http://wikimedia.de

Wikimedia Deutschland - Gesellschaft zur Förderung Freien Wissens e.V. Eingetragen im Vereinsregister des Amtsgerichts Berlin-Charlottenburg unter der Nummer 23855 B. Als gemeinnützig anerkannt durch das Finanzamt für Körperschaften I Berlin, Steuernummer 27/681/51985.