Re: comments on OWL test cases

On May 15, Jeremy Carroll writes:
> 
> ian:
> > Under these circumstances I'm not sure that the paragraph
>   my text
> > is necessary - the semantics of datatype theories was carefully
> > designed so that this kind of thing would NOT be necessary.
> 
> Hmm.
> If that's the case then I suspect we will end up aborting the LC vote.
> That understanding of datatyping gets into too many places for it to be simply 
> an editorial action to fix it.
> 
> Can I ask a concrete question:
> http://www.w3.org/2002/03owlt/editors-draft/draft/proposedByIssue#I5.8-003
> 
> _:a rdf:type owl:Restriction .
> first:p rdf:type owl:DatatypeProperty .
> first:p rdfs:range xsd:byte .
> first:p rdfs:range xsd:unsignedInt .
> _:a owl:onProperty first:p .
> _:a owl:cardinality "129"^^xsd:nonNegativeInteger  .
> first:john rdf:type _:a .
> 
> 
> this is inconsistent, but to know it you need to know about xsd:byte and 
> xsd:unsignedInt and their relationship.
> 
> What should a complete OWL Consistency Checker that only knows about integer 
> and string return?
> 
> My text says: Unknown
> Without it I think we have to say: Consistent, despite the test case being 
> inconsistent.

It is consistent if the interpretation of xsd:byte is just some subset
of the data value domain, which is the clearly defined semantics for
an unsupported datatype. This all seems perfectly clear to me, and
doesn't require any changes to any documents - certainly not an
abortion of the LC vote (and I would appreciate a general cooling of
the rhetoric regarding LC, the delay of which is a bogey man whose
appearance is now so regular as to be more boring than frightening).

Adding the condition that a Consistency Checker should return Unknown
in this case is a change to the design that I oppose, but don't feel
very strongly about. A compromise solution could be that checkers
should return a warning if ontologies contain unsupported datatypes.

Ian



> 
> Jeremy
> 
> 
> 
> 

Received on Thursday, 15 May 2003 11:57:46 UTC