Re: Joint Committee telecon today (semantics for domain and range)

From: Tim Berners-Lee ([email protected])
Date: 01/09/01


----- Original Message -----
From: "Peter F. Patel-Schneider" <[email protected]>
To: <[email protected]>
Cc: <[email protected]>; <[email protected]>
Sent: Tuesday, January 09, 2001 2:07 PM
Subject: Re: Joint Committee telecon today (semantics for domain and range)


>
> From: Dan Connolly <[email protected]>
> Subject: Re: Joint Committee telecon today (semantics for domain and
range)
> Date: Tue, 09 Jan 2001 12:59:06 -0600
>
> > "Peter F. Patel-Schneider" wrote:
> > >
> > > As promised, I have a new version of the semantics for DAML+OIL.  It
is
> > > enclosed below.
> > >
> > > The ONLY substantive changes are the last two lines, which read
> > >
> > >         <domain,?P,?C>  if <x,y> in IR(?P) then x in IC(?C)
> > >         <range,?P,?C>   if <x,y> in IR(?P) then y in IC(?C)
> > >
> > > NOTE:  This does not correspond to the current RDFS intended meaning
for
> > > these constructs.
> >
> > Intended by whom? ;-)
>
> I meant intended meaning in the sense that it is what is alluded to in the
> RDFS documentation, but not fully specified there.
>
> > It corresponds to what I intend, and at least one of the RDFS
> > editors (Guha). It does not correspond to the way some
> > folks have interpreted the RDFS spec.
> > I should double-check that this is on the RDF IG issues list...
> > no time just now. Please remind me if you get a chance...
>
> The RDFS documentation has a completely different meaning for multiple
> domains.  I know that there appears to be consensus to change RDFS in this
> manner, but I think that we need to attach caveats to every definition of
> domain and range until the change is made.  :-(

The specified RDFS meaning is in fact meaningless - the domain means
something along the lines that whatever the real:domain is, that this
rdfs:domain
is a subClass of it, in other words that there exists some class such that
all subjects of the predicate are in it and the given rdfs:domain value
is a subclass of it, which is always true as the Class Thing always has
these
characteristics.  So I don't see anyone defending the wording in the RDFS
spec.

tim


This archive was generated by hypermail 2.1.4 : 04/02/02 EST