W3C home > Mailing lists > Public > public-sw-meaning@w3.org > October 2003

Against Strong Ontological Commitment

From: Bijan Parsia <bparsia@isr.umd.edu>
Date: Fri, 10 Oct 2003 10:49:31 -0400
To: public-sw-meaning@w3.org
Message-Id: <F4EACDDD-FB30-11D7-9477-0003939E0B44@isr.umd.edu>

This is more of a test case, but it has immediate consequences for the  
WebOnt and perhaps the RDF working group.

***CLAIM: If Strong (or pt restricted) Ontological Commitment is true,  
then there can be no OWL DL RDF/XML documents*, all the syntax checkers  
are broken, and a lot of OWL Test manifests assert falsehoods.

*There may be some corner cases where the DL document doesn't use any  
of the OWL or RDF vocabulary. Hmm. :foo :bar "blargh". is an OWL DL  
document encodable in RDF/XML. So I'll change it to no "interesting"  
OWL DL RDF/XML document, where "interesting" means "using the rdf, rdf,  
or owl vocabularies. Actually, if you permit rdf entailment, it's true  
for all non-empty ones, I think.


Ontological Commitment requires (some form of) the imports closure of  
the used URIs in a document.

The owl document consisting solely of:
		<owl:Ontology rdf:about=""/>
Thus requires *at least* the importing of  
Thus, *at least* contains:

     <owl:imports rdf:resource="http://www.w3.org/2002/07/owl" />

(see http://mindswap.org/~bparsia/ontologies/test/importowl.owl)

This is in owl full due to redefinition of builtin vocabulary (bit more  


Ok, if you say, "Wait, owl:imports only has to have the *effect* of  
concatinating the graphs", well, you still, afaict, lose. I don't think  
the schema at http://www.w3.org/2002/07/owl is compatible with DL. At  
least the direct model theoretic semantics (in any straightforward way).

So there's lots of ways to fix this, change the definition of  
owl:imports, change stuff about species tests, change the document  
served up at http://www.w3.org/2002/07/owl (and the rdf: and rdfs:  
ones, perhaps).

I did think that some sort of RDDL document which said that the schema  
was "informative" and the spec "normative" might help. But who knows?

Bijan Parsia.
Received on Friday, 10 October 2003 10:51:38 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 19:56:01 UTC