- From: Holger Knublauch <holger@topquadrant.com>
- Date: Thu, 09 May 2013 10:25:19 +1000
- To: "public-vocabs@w3.org Org" <public-vocabs@w3.org>
On 5/9/2013 10:02, Dan Brickley wrote: > One of the trickier aspects of describing property semantics with OWL > is that we keep adding new terms into the evolving vocabulary. I am > wary of making strong claims in machine form today which would be > countered if, for example, we added a second or third type for some > property which currently only has one or two expected (ranges, or > domains) types listed. Yes I agree and understand that. But owl:inverseOf is hopefully a case that can still be stated strongly as it's unlikely that the same property will have different inverses in the future. Likewise, I hope the cardinality information is the same for all instances (owl:FunctionalProperty) or could otherwise be captured locally in the context of a class (mapping to owl:Restriction). --- Thinking out loud now: Regarding the rdfs:domains, in my personal opinion there are some reasons in favor of simply avoiding this scenario: just do not encourage property reuse, i.e. multiple domain classes for the same property. Take "location" as an example. It's currently allowed for Event and Organization. Apart from readability, I do not see reasons why the same property was reused here. It could just as well have been "eventLocation" and "headquarterLocation", possibly with an rdfs:subPropertyOf "location". Or introduce a shared superclass such as ThingWithLocation (not sure if multiple inheritance is supported here) and attach "location" to that. I know these things would be hard to change now, but maybe food for thought in the future evolution. Especially if you simply add new domains to a property such as "location" in the future, then existing applications may break because they have made certain assumptions about the old property. From an RDF point of view, I'd rather like to be able to efficiently query for schema:eventLocation than schema:location without also having to check the rdf:type of that thing. I am aware that there are arguments in favor of reusing properties, so please regard this only as one opinion among others. Cheers, Holger
Received on Thursday, 9 May 2013 00:26:02 UTC