Re: RDF Issue rdf-formal-semantics

From: Brian McBride <bwm@hplb.hpl.hp.com>
Subject: Re: RDF Issue rdf-formal-semantics
Date: Mon, 11 Mar 2002 16:05:30 +0000

> Hi Peter,
> 
> At 10:17 11/03/2002 -0500, Peter F. Patel-Schneider wrote:
> >Unfortuantely, the RDF model theory does not (yet) provide an acceptable
> >account of RDF semantics.  The current RDF model theory WD (of 14 February
> >2002) only provides an account for N-triples, and is missing other features
> >of RDF.
> >
> >In particular, there is no treatment of RDF reification or RDF containers
> >in the current RDF model theory.
> 
> Just so.  The Model Theory is not complete.  The key decisions on 
> reification and containers have been made however, and these will be 
> reflected in the next WD.

If so, then I await the next version of the model theory.

> The idea here is that we now have a mechanism, which was previously 
> lacking, for formally defining the semantics of RDF.  Defects in the model 
> theory can now be raised as specific issues against the model theory 
> document itself.  The overall catch all issue, that there is no formal 
> semantics for RDF(S), is resolved by the model theory document (the WG claims).

A formal semantics (only) gives meaning to a language.  Right now, the
correspondence between RDF syntax(es) and the model theory is not exact,
and thus the model theory is not a model theory of RDF.  Therefore, it is
not (yet) possible to answer (all) questions about RDF by appealing to the
model theory.

> >   The current RDF model theory also does
> >not consider datatypes or the new thinking on RDF literals.
> 
> To what new thinking are you referring?

Literals are apparently now triples, consisting of a binary tag, a string,
and a language tag.  The model theory only handles currently only handles
opaque literals.  (If these new literals are indeed opaque, i.e., equality
is exactly equality of the components, then the model theory still works.
Otherwise an addition is needed to the model theory.)

> >  Finally, the
> >translation from RDF/XML to N-triples is performed outside this model
> >theory, and this translation is quite complex.
> 
> Just so.  It is defined in the syntax document (also as yet incomplete):
> 
>    http://www.w3.org/TR/rdf-syntax-grammar/
> 
> Do you agree that the development of the model theory resolves the catch 
> all issue as you raised it, that there is no formal semantics for rdf and 
> rdfs.  There is now a semantics for RDF.  It is as yet, incomplete.

A workable model theory for all of RDF would resolve the issue, but the RDF
Core WG is not yet at that point, so the issue is not yet resolved.

> Brian

Peter F. Patel-Schneider
Bell Labs Research

Received on Monday, 11 March 2002 12:19:59 UTC