Re: RDF vocbulary compatibility

From: Jeremy Carroll <jjc@hplb.hpl.hp.com>
Subject: RDF vocbulary compatibility
Date: Fri, 07 Feb 2003 09:51:02 +0000

> 
> 
> I presented the current docs to an HP group, and was unable to explain why 
> the following RDF vocabulary is omitted from OWL Lite, for use with data:
> 
> Classes:
>   rdf:Bag
>   rdf:Seq
>   rdf:Alt
>   rdf:List
> 
> Properties:
>   rdf:_NNN
>   rdfs:member
>   rdf:first
>   rdf:rest
> 
> Other:
>   rdf:nil
> 
> As far as I can see there is no reason (other than aesthetics), although 
> incorporating rdf:_NNN and rdfs:member would involve only supporting their 
> use as ObjectProperty. Mixed usage, in which rdf containers are used with 
> both resource and literal values only fits in OWL Full.
> 
> The use of rdf:first and rdf:rest as part of the syntax for OWL does not 
> interact with their use as
> 
> rdf:value *does* seem technically problematic, as well as unaesthetic, 
> because it is primarily used with string values, but may be used with 
> resource values.
> 
> These could be added to OWL Lite, by treating the classes above as 
> owl:Classes,  the properties above as owl:ObjectProperties and rdf:nil as 
> an owl:Thing.
> Moreover, rdf:first and rdf:rest can be treated as owl:FunctionalProperty, 
> with domain rdf:List and the range of rdf:rest can be constrained to
>    rdf:List union { rdf:nil }.

RDF containers are extraordinarily problematic.  Their intended meaning is
not compatible with their formal meaning.  My preference would be to remove
them from OWL entirely.

> Moreover, I believe the current rationale for the existence of owl:Class as 
> a distinct concept from rdfs:Class is flawed.

Which rationale are you refering to?  

> In OWL Full they are the same concept.
> In OWL Lite and OWL DL any of the statements that might reveal differences 
> between the concepts are syntactically excised. Hence the justification 
> based on rdfs:Class including members such as owl:FunctionalProperty, 
> whereas owl:Class does not is spurious.
> 
> It would be simpler to just use rdfs:Class throughout, and nothing would 
> break. (The correspondence theorem may need additional care).

The RDFS semantics for OWL DL depends heavily on owl:Class, or, more
accurately on its class extension.  Removing owl:Class would be a major
change in the RDFS semantics for OWL DL.  

> Jeremy


peter

Received on Friday, 7 February 2003 17:05:18 UTC