- From: Jeremy Carroll <jjc@hplb.hpl.hp.com>
- Date: Mon, 10 Mar 2003 14:31:44 +0100
- To: <pfps@research.bell-labs.com>
- Cc: <www-archive@w3.org>, <schreiber@swi.psy.uva.nl>
Hi Peter, I have thought a bit more about the datatyping rule; and also realized we did not discuss one other aspect of difference (blankNode rdf:type rdfs:Class .) The datatype rule is: datatypeID => datatypeID rdf:type rdfs:Datatype . I think it will be easier to simply not permit rdfs:Datatype anywhere; and change the rule to: datatypeID => // empty Rationale: 1) The rule as is needs some change, since this should be at most optional for builtin datatypes (which are the only ones we have). 2) Making it optional then still leaves the orphaned datatype declaration problem; made worse since a datatype that is only used in a cardinality constraint gets orphaned. i.e. xsd:int rdf:type rdfs:Datatype . _:x rdf:type owl:Restriction . _:x owl:onProperty eg:p. eg:p rdf:type owl:DatatypeProperty . _:x owl:minCardinality 1^^xsd:int . is in OWL Full and not OWL DL, by the current mapping rules. 3) Two possible fixes to the orphaned datatype are: a) add new datatype declaration directive to Abstract Syntax (my preference) b) change mapping rule to not output any triple (easier) ===================== There are a number of places where the mapping rules include _:x rdf:type rdfs:Class . [opt] I have a mild preference for deleting them all; but leaving triples like classID rdf:type rdfs:Class . [opt] I requested the addition of such triples for uriref subjects for compatibility with RDFS; but I cannot think of a use case in which the corresponding triples for blank nodes aid compatibility. This is remembering: (a) blank nodes cannot be shared between documents and (b) RDFS does not use blank nodes as classes or properties (it is not actually prohibited, just not useful). Jeremy
Received on Monday, 10 March 2003 08:31:55 UTC