Re: owl:LiteOntology?

On Tue, 2002-10-22 at 19:53, Jim Hendler wrote:
> At 3:32 PM -0400 10/22/02, Mike Dean wrote:
> >I notice that we currently have no way of indicating to
> >tools (e.g. an "OWL Lite Validator" which checks for
> >sublanguage conformance) which level of the ontology
> >language the ontology author intends to be using.  As a
> >lightweight means of accomplishing this, I propose that we
> >define subclasses of owl:Ontology corresponding to each of
> >the language levels (with whatever names we eventually
> >choose).

Seems reasonable.

Do you expect it to have any
semantics impact? (beyond the obvious
fact that LiteOntology is a subclass
of Ontology). I hope not.

I wonder about test impact; the obvious test
case is something that says it's an owl:LiteOntology
but isn't. I'd put this in the same bucket
with some class with an rdfs:comment that
says it's infinite when it's not;
it's just "wrong" in a way that we don't formally
constrain; i.e. as far as test is concerned,
it's just an ontology that happens to not
meet the owl lite constraints.

> >	Mike
> Mike - how about instead of subclasses, we simply invent a "keyword" 
> field.  My reason for preferring this is that it then becomes 
> extensible

I don't understand this point; how is this extensible
in any way that owl:LiteOntology is not?

> - but also because if it will be easier for a tool to scan 
> for ontology declarations if it doesn't need to do inferencing to get 
> there.

I also don't understand this point; how are properties
(aka "fields") any different from classes when it
comes to doing inference versus syntactic scraping?

> I'd suggest something like
> <owl:ontology  owl:level="lite" rdf:about="">
>    <owl:imports .../>
>    <owl:version .../>
> </owl:ontology>
> If I've read the new RDF documents correctly, this is legal, clean 
> and an easy way to provide that mechanism.
>   -JH
> p.s. I can live with the subclassing, just find this somewhat 
> preferable as it will make things a little easier for my tool 
> builders.

Could you elaborate? I don't see how it's easier for toolbuilders
to look for a property value than a class.

Dan Connolly, W3C

Received on Friday, 25 October 2002 16:41:30 UTC