W3C home > Mailing lists > Public > www-rdf-interest@w3.org > October 2000

Re: definition of domain

From: Jeen Broekstra <jbroeks@cs.vu.nl>
Date: Thu, 12 Oct 2000 15:27:22 +0200 (CEST)
To: Tom Van Eetvelde <tom.van_eetvelde@alcatel.be>
cc: www-rdf-interest@w3.org
Message-ID: <Pine.GSO.4.20.0010121402040.17009-100000@flits.cs.vu.nl>
On Thu, 12 Oct 2000, Tom Van Eetvelde wrote:

> I clearly phrased my thoughts the wrong way. See answer to
> Pierre-Antoine. With applying the property to a subclass, I
> really mean the sublcass (node in a drawing) itself and not the
> subclass'es instances.

Ah, yes, I saw that after I had already sent my reply.
    
    <rdfs:Class rdf:ID="carnivore">
	 <rdfs:subClassOf rdf:resource = "#animal"/>
	 <eats rdf:resource="#animal"/>
    </rdfs:Class>

Actually I think it is an interesting idea. What you propose is
what is being done in OIL right now with slotconstraints:

    <rdfs:Class ID="carnivore">
	<rdfs:subClassOf rdf:resource="#animal"/>
	<oil:hasSlotConstraint>
	    <oil:ValueType>
		<oil:hasProperty rdf:resource="#eats"/>
		<oil:hasClass rdf:resource="#animal"/>
	    </oil:ValueType>
	</oil:hasSlotConstraint>
    </rdfs:Class>

The OIL way is more explicit, and allows other types of local
restrictions besides range restrictions. However, your way is
simpler and more intuitive.

As I mentioned earlier, this still does not allow you to infer
that a lion of class animal that eats gazelles is in fact a
carnivore. However, what you _can_ do is find inconsistencies,
such as carnivores that eat bananas. 

> This is exactly what my question is all about: why should this
> not be possible? Why has W3C put such restricitons on the use
> of 'domain'?  Do they have thorough reasons to define 'domain'
> in this way or did they just not see the use of applying
> properties not only to class instances, but also in the
> definitions of subclasses (if you draw this, you see that you
> apply the property then to the sublcass, hence my confusing
> terminology). I think W3C should change the semantics behind
> 'domain' in the RDFS spec, unless they have good design issues
> for the current semantics.

I suppose there are some unwanted consequences to what you
propose, one being that the interpretation of the use of a
property has to change when it is applied to a class instead of
an instance. But I think there are more RDF-knowledgeable people
here that can better answer this.

> It would be inefficient to push people to unintuitive modelling
> just because of some arbitrary definition in a spec. Sometimes
> we have to question the bible! Amen.

I definitely agree with you here, though I am not sure the RDFS
editors will appreciate their design decisions being labeled
"arbitrary" ;)

> By the way Jeen, I have to admit that your inferencing
> capabilities outrun mine. I can only argue that the example was
> ment to give an intuitive feeling of the direction I would like
> to go.

*grin* My "inferencing capabilities" come mostly from talking to
Description Logics people a lot. You either give up at some point
or you start understanding - a little ;)

Regards,

Jeen
-- 
                               Vrije Universiteit, Faculty of Sciences
Jeen Broekstra              Division of Mathematics & Computer Science
jbroeks@cs.vu.nl                                    de Boelelaan 1081a
http://www.cs.vu.nl/~jbroeks        1081 HV Amsterdam, the Netherlands
Received on Thursday, 12 October 2000 09:27:55 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:51:44 GMT