Re: LANG: Summary of Issues 5.6 and 5.14

From: Jeff Heflin <heflin@cse.lehigh.edu>
Subject: Re: LANG: Summary of Issues 5.6 and 5.14
Date: Thu, 19 Sep 2002 17:28:36 -0400

> Peter,
> 
> You said:
> 
> > Jeff's approach allows imports relationships between arbitrary ontologies.
> > Pat's approach only allows the importing of another ontology into the
> > current situation. 
> 
> But that's because Pat's approach is not fully specified. He said "If an
> ontology A contains [import B]." If we must express this in RDF, then
> the only way is with a triple "A imports B." I would be perfectly happy
> with using some other syntax than RDF, and not allowing imports between
> arbitrary ontologies, but I did not think that option was on the table.

No.  There are other possibilities.  For example, as someone suggested, you
could use triples like

      B rdf:type importMe .

or you could use my approach of having an attribute on the rdf:RDF element.

> You also said:
> 
> > Jeff's approach requires a notion of ontology in the
> > semantics, Pat's doesn't.
> 
> But Pat's does require a notion of ontology. What do you think A and B
> are? They are ontologies. 

Not really.  Pat's approach *might* require owl:ontology, but with null
semantics.  

> If you can't mention ontologies in the
> semantics, then you can't express his condition. 

No.  You could do the import at parse time.  There are other options
available as well.

> Note that one problem
> with Pat's expression is that it uses the symbol B as an identifier for
> the ontology and for the ontology itself. My graph() approach
> dereferences the symbol to get the ontology.

Well, this may be a desirable feature of Pat's approach.  Some ways of
doing his approach could dismiss with any notion of ontologies, except as
URI strings.

> Finally, you said:
> 
> > I think that the difference is, in fact, even greater.  Jeff's approach
> > appears to require semantic support.  Pat's approach can be done in the
> > syntax.
> 
> But we're discussing how we define imports in our documents, not how it
> can be implemented.  If we say, "If an ontology A contains [import B] (in
> whatever notation turns out to be appropriate) and if B + A entails C
> then A entails C" in our semantics for imports, then sure there's
> syntactic way of achieiving this effect, but I could implement it any
> way I wanted to.

I'm not talking about implementation.  I'm talking about specification
(i.e., how to define how imports works).  Implementors are free to
implement this however they want, as long as it gets the job done right.

> Jeff


peter

Received on Thursday, 19 September 2002 19:27:26 UTC