W3C home > Mailing lists > Public > public-owl-wg@w3.org > November 2007

Re: ISSUE-55 (owl:class): owl:class v. [sic] rdfs:class

From: Peter F. Patel-Schneider <pfps@research.bell-labs.com>
Date: Tue, 13 Nov 2007 03:49:18 -0500 (EST)
Message-Id: <20071113.034918.63329815.pfps@research.bell-labs.com>
To: alanruttenberg@gmail.com
Cc: public-owl-wg@w3.org

It probably is not a good idea to go to the OWL Reference document for
technical arguments.  Notwithstanding its name, the OWL Reference
document is not a normative docment nor does it provide complete
information.

In OWL DL, instances of owl:Class cannot be literals (e.g., strings and
numbers).  This semantic condition, along with some others, is given in
Section 5.4 of OWL S&AS http://www.w3.org/TR/owl-semantics/rdfs.html#5.4

It would seem, then, that OWL DL has some inferences that OWL Full does
not (because it has extra semantic conditions).  For example, in OWL DL
an OWL Class cannot have a string as a member, but this is possible in
OWL Full.  However, all these extra inferences involve syntax that is
not legal in OWL DL, so there is no way of seeing them in OWL DL.  (This
last assumes, of course, that Theorem 1 in
http://www.w3.org/TR/owl-semantics/rdfs.html#5.4 is correct.)

All this is a long-winded way of saying that there are difficult issues
to be probed if the DL-Full correspondence is to be maintained as much
as possible while conflating owl:Class and rdfs:Class.  However, if the
ontology is otherwise OWL DL and any consequences are also otherwise OWL
DL, and there is no direct probing of whether a class belongs to
owl:Class versus rdfs:Class, then it *might* be the case that rdfs:Class
can be turned into owl:Class without no visible differences.  This
remains to be proven, however.

Of course, this does not matter in the abstract syntax for OWL DL, nor
does it matter in the functional-style syntax for OWL 1.1, as there is
no use of either owl:Class or rdfs:Classes in these syntaxes.  This is
one of many reasons to prefer these syntaxes over an RDF syntax.

Peter F. Patel-Schneider
Bell Labs Research


From: Alan Ruttenberg <alanruttenberg@gmail.com>
Subject: Re: ISSUE-55 (owl:class): owl:class v. rdfs:class
Date: Mon, 12 Nov 2007 23:11:46 -0500

> Here is my understanding of why there is a distinction:
> 
> http://www.w3.org/TR/rdf-schema/#ch_class says:
> 
> "A class may be a member of its own class extension and may be an  
> instance of itself."
> 
> This is not the case in OWL-DL, where classes are not instances.
> 
> http://www.w3.org/TR/owl-ref/ says:
> 
> NOTE: owl:Class is defined as a subclass of rdfs:Class. The rationale  
> for having a separate OWL class construct lies in the restrictions on  
> OWL DL (and thus also on OWL Lite), which imply that not all RDFS  
> classes are legal OWL DL classes. In OWL Full these restrictions do  
> not exist and therefore owl:Class and rdfs:Class are equivalent in  
> OWL Full.
> 
> -Alan
> 
> On Nov 12, 2007, at 10:41 PM, OWL Working Group Issue Tracker wrote:
> >
> > ISSUE-55 (owl:class): owl:class v. rdfs:class
> >
> > http://www.w3.org/2007/OWL/tracker/issues/
> >
> > Raised by: James Hendler
> > On product:
> >
> > At ISWC 07, it has come up in several presentations that a great  
> > many OWL Full ontologies (in RDFS esp) become OWL DL if you just  
> > change rdfs:class to owl:class - while it is acknowledged that this  
> > makes a semantic change, many people said they simply just do it in  
> > their tools.
> >
> > Given the improved understanding at this point of things like  
> > punning, the WG might want to revisit this issue and see if there  
> > is a technical solution that could be considered - or at least make  
> > a clear and precise statement as to why we are unable to do so and  
> > why this dichotomy must consider to exist (as unfortunate as that is)
Received on Tuesday, 13 November 2007 09:02:25 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 8 January 2008 14:13:27 GMT