- From: Holger Knublauch <holger@topquadrant.com>
- Date: Thu, 16 Mar 2017 15:28:21 +1000
- To: public-data-shapes-wg@w3.org
Hi Dean, thanks for that. I am beginning to have second thoughts about this whole topic and believe our response should be to do nothing at all (in the spec). Yes, Protege currently may have problems handling SHACL files, but that's nothing that we can address. I don't want to comment on the fact that OWL 2 has moved away from RDF apart from saying that it did have some long-term effects to the interoperability with RDF-based languages. And that it's clear that we cannot expect tools primarily built for OWL to make "sense" of SHACL - how and why could they anticipate this? We have tried to make life easier for RDFS/OWL-based tools, e.g. by adding rdfs:domain and range statements to the SHACL namespace so that these tools understand which properties are typically used for certain resources, and which values they should have. Otherwise we don't expect any harm and SHACL will typically still work even if these triples are used for inferences. Note that we do use owl:imports as the only URI from the OWL namespace. We furthermore do recommend that sh:prefix triples be attached to the "graph" resource of a graph, which often is declared to be an instance of owl:Ontology. Some tools (including ours) use the X rdf:type owl:Ontology triple to detect the graph URI, so I would prefer to keep that triple in for now. This doesn't really add a dependency on OWL in my opinion. Which leaves us with the issue on how to explain to users of Protege that they may have to wait for a SHACL plugin or at least for a locally adjusted version of the SHACL TTL file which makes certain properties a owl:AnnotationProperty. But this temporary situation is not a topic for a specification. I would just like to respond to Dave's comment along these lines and move on. Holger On 16/03/2017 1:05, Dean Allemang wrote: > I made a draft of the usage note in response to Dave McComb's comment, > and I wrote to him to socialize that. > > Here is my draft. It is quite short, and modeled on the RDFS > paragraph that's already there. Since we *aren't* using OWL > inferencing, it is a lot shorter. > > > A related topic - the only reference to OWL that we make in shacl.ttl > is to owl:Ontology. Is there a way to make SHACL completely > non-dependent on OWL? Do we even need to declare an owl:Ontology in > shacl.ttl? Part of the motivation for not declaring things as > annotation properties is that we aren't using OWL. And we really > aren't; we never actually define any entity that is a member of a > class in the OWL namespace (except the one owl:Ontology at the top). > > > Dean
Received on Thursday, 16 March 2017 05:29:02 UTC