Re: Proposed response to Golbeck regarding imports issue

From: "Jeremy Carroll" <jjc@hplb.hpl.hp.com>
Subject: RE: Proposed response to Golbeck regarding imports issue
Date: Tue, 17 Jun 2003 10:29:54 +0200

> > > continued for a long time after.  The decision recorded in our
> > > records is in:
> > > http://lists.w3.org/Archives/Public/www-webont-wg/2002Nov/0004.html
> > > and you will see it says nothing about whether a namespace reference
> > > (or other naming) does or does not import.
> >
> > My reading of this message is quite different.  The document exactly
> > defines which other documents are to be considered to be
> > imported, and this
> > set of documents is determined solely from imports directives.
> >
> ..
> > peter
> 
> I would balance this by, while OWL software must be able to strictly follow
> the recommendation, it seems quite plausible to me for OWL software to have
> additional features that monotonically extend the recommendation. 

Sure, it is plausible that software might extend the recommendation, and
that some extensions might be better than others.  However, I do not think
that our documents should anywhere mention or otherwise even hint of
condoning this.  Such extensions are *not* OWL.

Our hands are actually somewhat more tied than they could be because of the
requirement that OWL use RDF graphs, and the common feeling that all
information in the Semantic Web conveyed using RDF graphs.  If this
situation did not hold, then it would be easier to true extensions to OWL,
by using extra syntax.  However, in the current situation extensions to OWL
necessarily turn OWL non-entailments into entailments, just as extensions
to RDF (including OWL) turn RDF non-entailments into entailments.  

> However,
> these will not interoperate. So I would not be displeased with a user option
> to import all the definitions of all terms I used (if the sw devleoper can
> come to an acceptable definition of 'definition'), nor would I be displeased
> to have a user option to import all the namespaces declared.

Sure, as long as there is no indication from the working group that this is
an any way related to OWL.

> In certain contexts, I think OWL would be useful with some ontologies
> preimported.

Well, this would not be OWL.  

> Such user options are not what we are currently trying to work on.

I very much agree with this.

> Jeremy

peter

Received on Tuesday, 17 June 2003 07:09:20 UTC