Re: is rdfs:domain useful as currently defined?


First thought:

The practical alternative for representing a "polymorphic" property would 
seem to be to introduce a new class to represent the domain, and to define 
it to be a subclass of the various required classes.  Doesn't that take you 
back to the same position?

Second thought:

Does it matter?  Is some valuable inferencing capability being lost?  I 
have a hunch that any form of general (non-closed-world) RDF inferencing 
that depends on knowing everything about something is doomed to fail.  (Why 
am I getting a sense that context is important here?  Mumble.)


At 04:27 PM 6/6/00 -0400, Ralph R. Swick wrote:
>As we've been studying the uses of the RDF Schema vocabulary for
>reasoning about models described in RDF, we've come to realize
>that the defined 'domain' property is not particularly useful
>for inferencing:
>  "If there is more than one domain property, the constrained
>   property can be used with instances of <em>any</em> of the classes"
>Since we are working in a non-closed world, we can never know whether
>we have all the possible domain statements that might apply to a
>property so we can never compute a definitive list of the possible
>classes of the subject of a statement with this property as predicate.
>At best, the domain property we've defined permits determination
>that no known constraints have been violated.  This is what the
>Working Group intended as far as I can tell, largely at my own
>recommendation.  But I'm having second thoughts.
>I haven't had a chance to examine other implementation work in
>detail to see how people have used rdfs:domain.  At a minimum,
>it might be appropriate to change its name so that it is more
>clearly distinguished from rdfs:range which *does* allow inferencing.

Graham Klyne

Received on Wednesday, 7 June 2000 05:35:13 UTC