Re: Proposal and Test cases (Re: skolems: visible differences?)

To go a bit further in this vein, it's pretty clear that the  
semantics are primary, not the conformance levels. I presume that the  
reason we have entailment tests, for example, in Old Owl is to test  
whether the semantics have been respected. Given a sensible  
semantics, the need for speccing conformance per se for entailment is  
lower since we have known definitions. You respect the semantics by  
having your entailment services respect that semantics.

Furthermore, there is support (e.g., Sandro voiced it at the F2F and  
I saw nods; Peter just voiced it; I hereby voice it) for supporting  
conformance levels to help interoperability of systems in what they  
actually do for users and help guide implementors.

Core services for OWL DL users are classification and realization,  
both forms of entailment. That's *the point* of selecting ontology  
languages like OWL (e.g., see Alan Rector's very illuminating  
presentation: <http://rease.semanticweb.org/ubp>). Future query  
languages (which are coming soon) involve entailment.

How does it help mitigate user surprise for them to load an ontology  
up, get the same consistency results, then classify and get different  
class hierarchies from Pellet and from FaCT++? "Oh, well, y'know, we  
conform to *owl* which is designed so that it doesn't say anything  
about entailment, sorry."

It's hard to see how we can make conformance games work for this  
issue. Usually, you muck with conformance in order to help  
implementors, not to help spec writing (e.g., making something  
optional if some system couldn't reasonably adapt to i t).

On the other hand, if you don't care about entailment, it's hard to  
see why you think it's important to maintain variable *semantics*.  
Why not change OWL Full to skolem semantics as well? It would  
arguably simplify the semantics presentation.

Cheers,
Bijan.

Received on Thursday, 17 January 2008 16:56:07 UTC