Re: a question regarding to semantics of rdfs:subClassOf

From: Jos de Bruijn <jos.debruijn@deri.org>
Subject: Re: a question regarding to semantics of rdfs:subClassOf
Date: Wed, 12 Oct 2005 10:46:20 +0200

> Hi Alan, all,
> 
> > Can someone please clarify this for me.
> > 
> > In Section 3.2.1 of OWL Web Ontology Lang Ref, it mentioned that
> >         The rdfs:subClassOf construct is defined as part of RDF
> >         Schema. Its 
> >         meaning in OWL is *exactly the same* .....
> > However, RDFS semantic conditions table in Section 4.1 of RDF
> > Semantics 
> > defines rdfs:subClassOf to be an "if ... then ... " relationship,
> > whereas the If-and-only-if conditions table in Section 5.2 of OWL
> > Semantics 
> > and Abstract Syntax defines rdfs:subClassOf to be "iff" 
> > 
> > From the definition, it seems that OWL defines subclass relationship
> > stronger than RDFS does. My intuitive understanding (please correct me
> > if I am wrong)
> > is that, in OWL, if the class extension of c1 is a subset of class
> > extension
> > of c2, then c1 is subClassOf c2.
> 
> You are completely right. The subclass relationship in OWL is indeed
> stronger than the subclass relationship in RDFS. I guess this is a
> mistake in the OWL Reference document (I'm CCing public-webont-comments,
> hoping this mistake will be rectified in the errata).

In defense of the document, it was probably correct at the time it was written.
At one time rdfs:subclassOf had an extensional meaning, i.e., the same as the
OWL meaning, but the RDFS meaning was later changed.

> The authors of the OWL reference document may have been misled by the
> informative section 4.2 of the RDF semantics document [1] which
> describes a possible extension of the RDFS semantics to include the
> if-and-only-if definitions which are in OWL.

I do not believe that this is the case.  In particular, I believe that this
wording was not part of the RDF Semantics document when the OWL Reference
document was written.

Of course, the change to the RDF model theory should have triggered a change to
this part of the OWL Reference document.  In any case, I believe that an
erratum should be issued.

> Best, Jos
> 
> [1] http://www.w3.org/TR/rdf-mt/
> 
> > 
> > I am not trying to be picky about the wording here. Just try to
> > understand
> > this better.
> > 
> > Thanks,
> > 
> > Zhe (Alan) Wu
> > Oracle

Peter F. Patel-Schneider
Bell Labs Research

Received on Wednesday, 12 October 2005 13:12:05 UTC