- From: Bo Ferri <zazi@smiy.org>
- Date: Sat, 18 May 2013 01:40:19 +0200
- To: public-vocabs@w3.org
Hi Holger, On 5/18/2013 12:50 AM, Holger Knublauch wrote: > > On May 18, 2013, at 8:42 AM, Bo Ferri wrote: > >> Hi Holger, >> >> this looks pretty good so far. Thanks a lot for your engagement. However, I'm not sure whether your utilisation of rdf:Property and owl:unionOf ranges that include classes as well as datatypes are really possible. Albeit, from the OWL 2 specs [1] I can see that there are definitions for ObjectUnionOf and DataUnionOf, which are somehow separate from each other, or? For example rNews (e.g. [2]) makes use of object unionOfs and datatype unionOfs, but no mixing. > > Yes these problems (among others) caused me to remove the mixed properties in my first version. It is IMHO *very* unfortunate that the schema.org design uses such mixed properties and I hope this pattern will not be propagated further. Instead, new properties should be introduced for the string values. But given this pre-existing design, a compromise needs to be applied. And here the compromise is to use OWL Full over OWL DL. Generally, I would like to see such mixed ranges. I just wasn't really sure whether this could be possible. However, when I interpret your comment correct, then it is really possible and valid. > > >> Generally, you could add rdf:Property to all properties and rdfs:Class to all classes for "backward" compatibility (this is one of my personal preferences when modelling ontologies ;) ). > > I personally don't like having redundant triples in a file. Anyone can quite easily add those (either dynamically as "inference" or through a CONSTRUCT query) to their own copy of the schema file. Yes, I know. However, I think that this should be part of the common mapping rather then the consumer has to do it on his/her own. Anyway, I think that this is only a minor issue that could be argued for both perspectives ;) So I would really vote for your new proposal (with the object + datatype unionOf ranges and the utilisation of rdf:Property) and that this mapping should be publish at schema.org. Cheers, Bo
Received on Friday, 17 May 2013 23:40:55 UTC