Re: skolems: visible differences?

On Jan 13, 2008, at 2:03 PM, Bijan Parsia wrote:

> On Jan 13, 2008, at 1:23 PM, Peter F. Patel-Schneider wrote:
> [snip]
>> Hmm, but this is not complete with respect to entailment in the RDF
>> semantics, as far as I can tell.
> [snip]
>
> It is, in fact, not complete.
>
> I regard that fact as a feature, not a bug. It is more important to  
> be backward compatible with universal practice, expectation, etc.  
> than with a (IMHO) buggy (rather, mal-designed) prior spec.
[snip]

I realized that though I sincerely didn't mean it that way, this very  
easily read as in a highly inflammatory tone.

I certainly don't mean to suggest that it was obvious at the time of  
its design that it was broken. At the time, I certainly thought it  
was a reasonable, even sensible, and 'clean' approach. My subsequent  
experience with it both in the rdf and owl worlds was that it is  
entirely wrong. It is simply ignored by users and implementors, and  
specs like SPAQRL and RIF ignore that bit as well. It causes  
theoretical problem (with practical impact) at even the RDF/RDFS  
level) etc. etc.

Hence, I feel like we shouldn't be strongly bound by the letter of  
the spec. I believe treating BNodes as skolem constants will enhance  
the interoperability of RDF and OWL systems and be the least  
surprising for users and that these considerations are strong enough  
to override the requirement for spec compatibility.

Cheers,
Bijan.

Received on Sunday, 13 January 2008 14:12:45 UTC