- From: Patrick Stickler <patrick.stickler@nokia.com>
- Date: Mon, 11 Aug 2003 10:14:24 +0300
- To: <w3c-rdfcore-wg@w3.org>
I like the idea proposed by Frank during the last conference call that insofar as the MT is concerned, the value space of rdfs:XMLLiteral is disjunct from all XSD value spaces, and from the set of plain literals, *BUT* that we somewhere note that one may apply regular (1:1), bidirectional "conversions" (or "coercions") between the members of the value spaces of particular pairs of datatypes. Thus, while "1"^^xsd:integer and "1.0"^^xsd:float are semantically distinct values and are not equal, one may convert between them (coerce the value from one datatype to the other), back and forth, as often as one likes, without any loss of meaning. Likewise, while "foo"^^rdfs:XMLLiteral and "foo" (a plain literal without lang tag) and "foo"^^xsd:string are all semantically distinct values, one may convert between them freely. In fact, we can expect RDF applications to do this frequently. I think this distinction between equality versus regular bidirectional convertability addresses the two previously (thought to be conflicting) needs of keeping the MT sane (by keeping value spaces disjunct) and yet capturing the common intuition that certain datatypes still have a kind of 1:1 intersection of values. That intersection is thus not equality, but 1:1 bidirectional convertability. I think we can use this idea of convertability to finesse solutions to the remaining issues regarding the relationships between XML literals, plain literals, xsd:string, and in the future, between XML literals and XSD complex types. Knowing that such 1:1 conversion relations can be defined to hold between datatypes alleviates my concerns about defining value spaces as disjunct. Patrick
Received on Monday, 11 August 2003 03:14:35 UTC