Re: Denotation of owl:Class

Jeremy:
>>i.e. globally replace owl:Class by rdfs:Class in OWL S&AS and 
>>everything still works.
Pat:
>Except that there would then be entailments that weren't legal in OWL-DL.

I overstated the ease of the replacement, there would be some consequential 
changes needed in the correspondence theorem, for instance.

>>You cannot say owl:Class rdfs:subClassOf rdfs:Class or the opposite in OWL
>>Lite and OWL DL so the need for this distinction is moot.

>But the point is that OWL-DL  needs a name for its universe that it 
>can use. It can't define it, and probably OWL-Full can't either, in 
>fact.

OWL DL (direct semantics) can quite happily call its universe the universe, or 
the domain of discourse or whatever. owl:Class only appears in the first 
sections of S&AS in the mapping rules. I ask the question, what would have to 
change in the RDFS compatible semantics to accomodate changing the mapping 
rules by replaceing owl:Class with rdfs:Class.

OWL DL (rdfs compatible semantics) does need a concept corresponding to 
owl:Class (i.e. some subclass of rdfs:Class, not including the builtin 
classes), but it does not need to give it a URI. The only possible problems 
with using the uri rdfs:Class for the concept currently called 'owl:Class' in 
OWL DL are eliminated by the syntactic restrictions.

I would like you to exhibit a test case of an entailment that would hold in 
OWL Full, but not in OWL DL, which is not excluded by the preamble to the 
correspondence theorem.

I would agree this is a delicate point, and don't particularly want to press 
it; and I suspect that I would need to do quite a lot of work (more than I am 
prepared to) in order to convince even one WG of the correctness of my 
position.

Jeremy

Received on Saturday, 26 April 2003 15:38:14 UTC