Re: ACTION-178: What is ISSUE-116 (Axiomatic Triples for OWL R) about?

From: Alan Ruttenberg <alanruttenberg@gmail.com>
Subject: Re: ACTION-178: What is ISSUE-116 (Axiomatic Triples for OWL R) about?
Date: Tue, 12 Aug 2008 13:22:14 -0400

> 
> On Aug 12, 2008, at 11:31 AM, Peter F. Patel-Schneider wrote:
> 
> >
> > From: Alan Ruttenberg <alanruttenberg@gmail.com>
> > Subject: Re: ACTION-178: What is ISSUE-116 (Axiomatic Triples for OWL R) about?
> > Date: Tue, 12 Aug 2008 11:22:40 -0400
> >
> >> On Aug 12, 2008, at 11:16 AM, Ivan Herman wrote:
> >>
> >>> Alan Ruttenberg wrote:
> >>>> On Aug 12, 2008, at 11:02 AM, Michael Schneider wrote:
> >>>>> Alan Ruttenberg wrote:
> >>>>>
> >>>>>> Is the assumption that in OWL R/Full that the only way in which one
> >>>>>> determine entailments is to forward chain rules and then look in
> >>>>>> the resultant triples for it?
> >>>>>
> >>>>> It's not an "execution semantics". Instead, there has to *exist* a
> >>>>> finite sequence of rule applications, which leads from the LHS to the RHS
> >>>>> (or an upper graph of the RHS).
> >>>> OK. I'm then confused by Ian's comment about generating infinite
> >>>> numbers of triples:
> >>>
> >>> But his comment _was_ justified. In
> >>>
> >>> http://www.w3.org/TR/rdf-mt/#RDFINTERP
> >>>
> >>> the number of axiomatic triples is infinite indeed, due to the rdf:_i
> >>> terms. Herman ter Horst's trick of keeping it finite for a specific
> >>> graph tries to get around that problem and thereby making it possible to
> >>> do, eg, a forward chaining on a specific graph (with the caveat that
> >>> PFPS just referred to in another mail:-(
> >>
> >> Yes, but my question to Michael was to this point.  From what he says I
> >> don't see that the theoretical existence of an infinite number of
> >> triples necessitates their materialization by forward chaining. There
> >> are other methods of using rules.
> >>
> >> -Alan
> >
> > Which of the other methods of using rules are germane to this
> > discussion?
> 
> I am trying to understand the underlying assumptions in this
> discussion. If no other method of rule applications than forward
> chaining are being considered and a solution to this issue should assume
> a forward chaining rule engine that would be useful information. What
> prompts the questions and comment were an earlier discussion in which it
> was mentioned that JENA used a combination of forward and backwards
> chaining.
> 
> Therefore your question is one I would I would like answered as well,
> but I am not the one to answer it.
> 
> -Alan

I don't see that any of the other methods for using rules are even
potentially germane to this discussion, regardless of what JENA does.

peter

Received on Tuesday, 12 August 2008 17:47:36 UTC