bug in RDF model theory having to do with rdfs:subClassOf and rdfs:subPropertyOf

The RDF model theory document of 29 April 2002
(http://www.w3.org/TR/2002/WD-rdf-mt-20020429/) has a bug having to do with
rdfs:subClassOf and rdfs:subPropertyOf.

In the model theory it is not necessary for rdfs:subClassOf or
rdfs:subPropertyOf to be transitive in rdfs interpretations.  This goes
against the clear intent of RDFS 1.0
(http://www.w3.org/TR/2000/CR-rdf-schema-20000327/).  It also means that
RDFS closures do not correspond to RDFS entailment, as closure rules rdfs5
and rdfs8 are invalid.

There are two ways of fixing this bug.  If complete compatability with RDFS
1.0 is needed, there should be separate semantic conditions on RDFS
interpretations that make rdfs:subClassOf and rdfs:subPropertyOf
transitive.  It would also be possible to turn the existing semantic
conditions for rdfs:subClassOf and rdfs:subPropertyOf into iff conditions,
but this would be a change from RDFS 1.0, although I can't think of any
RDFS-observable consequences between this fix and the previous fix.

I note in passing that making the semantic conditions for rdfs:domain and
rdfs:range into iff conditions *would* have RDFS-observable consequences,
even if the semantic conditions for rdfs:subClassOf and rdfs:subPropertyOf
are not changed.

Peter F. Patel-Schneider
Bell Labs Research

Received on Friday, 20 September 2002 14:28:07 UTC