W3C home > Mailing lists > Public > w3c-rdfcore-wg@w3.org > September 2001

Re: model theory publication draft

From: Graham Klyne <Graham.Klyne@MIMEsweeper.com>
Date: Thu, 20 Sep 2001 04:37:59 +0100
Message-Id: <>
To: Pat Hayes <phayes@ai.uwf.edu>
Cc: w3c-rdfcore-wg@w3.org
At 03:58 PM 9/19/01 -0500, Pat Hayes wrote:
>THIS is the case I originally had in there but decided to remove. It is 
>the analogous inference for domains:
>aaa rdf:domain zzz .
>zzz rdfs:subClassOf yyy
>aaa rdf:domain yyy
>and similarly for rdf:range, of course.

My intuition is that this is not really helpful.

I think the use of rdf:domain (etc.) comes from situations like this:

   aaa rdf:domain zzz .
   ppp aaa qqq .


   ppp rdf:type zzz .

which, together with with

   zzz rdfs:subClassOf yyy


   ppp rdf:type yyy .

which I think is the "useful" conclusion one might draw from:

   aaa rdf:domain yyy .

It's not clear to me that being able to deduce superclass domains really 
provides any useful information, and might increase the workload of an 
inference engine.

This view is reinforced by the following consideration. A review of the 
schema-closure rules shows that the only triple one can generate from an 
rdfs:domain statement is an rdf:type statement.  The above demonstration 
suggests that deducing superclass domains doesn't allow one to deduce any 
rdf:type statements that were not already available.

In summary, I think, in the absence of specific contrary requirements, to 
stick with 'if' rather than 'iff' for the domain (and range) conditions.


Graham Klyne                    MIMEsweeper Group
Strategic Research              <http://www.mimesweeper.com>
Received on Thursday, 20 September 2001 01:22:02 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 20:24:04 UTC