Re: Rules WG -- draft charter -- NAF

From: Stefan Decker <stefan@ISI.EDU>
Subject: Re: Rules WG -- draft charter -- NAF
Date: Tue, 18 Nov 2003 17:56:10 +0000

> Peter,
> 
> At 05:33 PM 11/18/2003, Peter F. Patel-Schneider wrote:
> >From: Stefan Decker <stefan@ISI.EDU>
> >Subject: Re: Rules WG -- draft charter -- NAF
> >Date: Tue, 18 Nov 2003 17:13:36 +0000

[...]

> > > >Where did Description Logics come into the story?  Just about any language
> > > >that goes beyond ground atomic facts and datalog has this issue.  How 
> > would
> > > >you close P(a) v P(b) with respect to P?  Even just adding functional
> > > >properties to RDFS causes issues.
> > >
> > > Maybe we should start slowly by defining exactly what the problem is 
> > that the
> > > rule language is supposed to solve, and then we may try to look for 
> > solutions.
> > > It could be a good start if we are just defining rules for RDF graphs and
> > > not take the
> > > semantics of the language encoded in the RDF graph into account.
> >
> >Huh?  How is this going to get anywhere?  Why would anyone care about a
> >rule system that acts on the syntax of RDF (or, for that matter, on the
> >syntax of any logic)?  I see no benefit whatsoever along this path.
> 
> You are associating RDF mostly with logic, many (I would claim most)
> people use RDF is to exchange  data, which needs to be transformed or queried.
> There a rule language would make a lot of sense.

Huh?  How are mediums for the exchange of data significantly different from
logics?  Do you really mean to have rules work on the syntax of RDF?  

To me, a rule that works on the syntax of RDF would be something like:

if   <rdf:Description rdf:type="?x" ??a>??y</rdf:Description>
then <?x ??a>??y</?x> 

which converts some RDF/XML syntax into a shorthand form with the same
meaning.

> > > If we want to take the semantics into account, let us try to define 
> > what the
> > > problem is, and look at which part of the problem a rule language can
> > > solve, and
> > > how to incorporate solutions for the other problems.
> > > Do you have a definition of the problem you are trying to solve?
> >
> >No, but I'm not trying to restrict the possible solutions, either.  All
> >I've been saying is that CWA and other circumscriptive notions have a
> >decided cost in very many settings.
> Definitely - if we talk about circumscription for FOL theories.
> Non-monotonic operators for datalog are much cheaper.
> Lets find out what we want.

Agreed, but then pronouncements like ``CWA is easy'' should not be allowed
because we don't yet know whether the problem will admit a solution where
CWA is indeed easy.

> Best,
>          Stefan

peter

Received on Tuesday, 18 November 2003 13:05:39 UTC