- From: Markus Lanthaler <markus.lanthaler@gmx.net>
- Date: Tue, 11 Feb 2014 21:04:10 +0100
- To: <public-hydra@w3.org>
On Tuesday, February 11, 2014 6:53 PM, Gregg Kellogg wrote: > On Feb 11, 2014, at 1:09 AM, Ruben Verborgh wrote: > > > HI Gregg, > > > >> foaf:Person rdfs:subClassOf owl:Restriction [ owl:onProperty > foaf:familyName; owl:cardinality "1"^^xsd:unsignedInteger ] . > >> > >> This certainly makes the presence of familyName required, although > it also has the effect of making it required on all subclasses of > foaf:Person I believe. > > > > The problem is that this changes the foaf:Person class; you don't > want to do that. > > Yes, that's true, but the alternative mechanism using declaring a > hydra:supportedProperty is also making a universal claim on > foaf:Person, which could be interpreted as meaning that any service > having a foaf:Person class also has a such a supported property, > although it doesn't have the same entailment consequences. Yeah, that's probably also we want to discuss and clarify in the spec. The alternative is to use (anonymous) subclasses which specify the supported properties as Sam did in his demos. > From a practical point of view, if either claim is made in a service- > specific vocabulary the potential for such claims to "infect" outside > uses is limited. > > This also puts into question _any_ use of OWL to make claims that are > service specific. In any case, a hydra:required property does overlap > cardinality constraints, and if it's appropriate to duplicate such > functionality in a Hydra vocabulary, perhaps it should rely on the same > concepts. You mean use the same model as OWL restrictions? The more or less only reason to introduce "required" is to drastically simplify its usage. Expressing OWL restrictions correctly is, for most people, not that trivial. We were discussing to use a little bit of OWL to make the description of Hydra itself more explicit. This could then be used to render better documentation. Users of Hydra, however, neither need to understand OWL nor know that we use it :-) If there's interest, we can always create follow-up documents which describe more sophisticated modeling using OWL etc. -- Markus Lanthaler @markuslanthaler
Received on Tuesday, 11 February 2014 20:04:42 UTC