RE: The Peter paradox isn't.

This might be a start at resolving the problem.  However, you have to be
very careful to not allow the creation of something related to rdf:type.
One avenue, which you have forbidden below is via rdfs:subPropertyOf.

In essence, you are trying to wall off rdf:type from all (other)
properties.  If this is done successfully, then there should be no (or no
useful) observable difference from eliminating rdf:type from
interpretations entirely, and using only CEXT.  (This is closer to the
description logic way of providing typing.)

It may be that this works, but there may also be other things that have to
be similarly walled off.  One candidate for this would be rdfs:subClassOf.

peter


From: "Jeremy Carroll" <jjc@hplb.hpl.hp.com>
Subject: RE: The Peter paradox isn't.
Date: Thu, 21 Feb 2002 13:09:26 -0000

> > > I had wondered whether changing (a) and not having rdf:type as 
> > a property
> > > would be the simplest fix.
> > 
> > I am not sure what the implications of "changing (a)" are, but 
> > rdf:type has
> > a central role in RDF 1, specifically for the typedNode 
> > production. There is
> > much software which depends on this, including software that traverses
> > rdf:type arcs as part of the inferencing process.
> > 
> > The current RDF work is not intended to 'break' current software.
> > 
> 
> The change I was imagining is that:
> - typed node production continues unchanged,
> - triples involving rdf:type continue unchanged
> But
> 
> rdf:type rdf:type rdfs:Property .
> 
> is false.
> 
> So that the rdf document (namespace liberties)
> 
> <rdf:RDF>
>   <rdfs:Property rdf:about="rdf:type" />
> </rdf:RDF>
> 
> is a contradicition.
> 
> And any use of rdf:type with rdfs:subPropertyOf is also a contradiction.
> 
> i.e. a change to RDF Schema not to RDF M&S.
> 
> Certainly if these changes impact deployed code then they are not minor.
> 
> Jeremy
> 
>

Received on Thursday, 21 February 2002 09:47:46 UTC