Re: SEM: light reviewing of OWL ASS document

From: "Jos De_Roo" <jos.deroo.jd@belgium.agfa.com>
Subject: SEM: light reviewing of OWL ASS document [Re: semantics document moved]
Date: Tue, 5 Nov 2002 02:18:15 +0100

> I think I (at last) understand the document...
> It seems as if there is a lot inspired by sound,
> complete and efficient reasoners more specifically
> DL reasoners (e.g. 3 times that side condition that
> properties that are transitive...).

Well, not really.  A model theory does give a complete characterization, by
definition.  A reasoner can then be sound and/or complete with respect to
the model theory.  

On the other hand, the side condition on transitive properties *is* there
so that complete reasoning for OWL/DL is decidable as is the division
between individuals, classes, and properties.

> On the other hand it seems to me that there are also
> other decidable fragments of OWL Full (thinking about
> Prolog's for instance) so why not at least mention
> that (and it's indeed another big job to work out the
> details to the same level as OWL DL, but why not?).

Well, perhaps someone could do it.  I'm not very motivated to do so, and
certainly don't have time for it in the next ten days or so.

> Datatyping seems to converge with RDFCore work, which
> is good news, just that I see owl:Datatype; is that
> also co-extensional with rdfs:Datatype in OWL Full?

The Datatype stuff is still unfinished, waiting on the details of Pat's
model theory for RDF datatyping.  I won't be able to determine whether
owl:Datatype and rdfs:Datatype are the same in all styles of OWL until I
see the RDF datatyping model theory.  I do expect that in OWL/Full the two
will line up.

> Use of the terms OWL Lite, OWL DL, OWL Full is not
> always there (e.g. fast OWL is there 5 times and
> OWL as such also a couple of times) but I guess
> that can be fixed easily.

Yes, I've been working on this.  No more Fast or Large OWL.  OWL by itself
is only used when talking about OWL in general.

> Typo in 3.4    ... is llowed
> also near the end of the 4. table there are 2 triples
>   <URI reference> <URI reference>)
>   <URI reference> "<lexical-form>")
> 
> I'm still thinking that more oneway IF's in 5.2
> are feasable (e.g. for domain and range).

Making domain and range ifs instead of iffs may be doable, but does cause
some potential problems.  I think that it could work out, but there remains
work to be done.

> In 5.1 VRDFS is not including rdfs:Literal and rdf:nil
> and in Appendix A VRDFS in not including rdfs:Literal
> Why different? Is it a good idea to explain why these
> exclusions are there in VRDFS and VOWL?

The reason, and perhaps this should be better explained, is that
the excluded names can be mentioned in OWL/DL.  I'll try to see if I can make
VRDFS be the same in both 5.1 and Appendix A.  

> Jos De Roo, AGFA http://www.agfa.com/w3c/jdroo/


peter

Received on Monday, 4 November 2002 21:01:34 UTC