- From: Alain Michard <Alain.Michard@inria.fr>
- Date: Wed, 05 Aug 1998 10:50:54 +0200
- To: www-rdf-comments@w3.org
Hello, We're trying to design a RDF-based information retrieval system, and we discover that we really need to define a few general-purpose Classes, as possible values of RDFS:Range The most obviously needed are (non exhaustive list!) : Numerals Integers PositiveIntegers URI MimeType Of course we can perfectly define these Classes in our own schema, and implement in our application the corresponding processes (validity checking of declarations, automatic computation of value-ranges by transitive closures, etc.) I'm concerned by the interoperability problems that may rapidly emerge if many developpers define their own schema with these basic classes locally re-defined. Some systems will process foo:Integers, others Bar:Integers, and how would they recognise ++efficiently++ that when some declarations are imported from system A to system B, foo:Integers is (hopefully!) strictly equivalent to bar:Integers ? The rdf and rdfs constructors are of course "wired" in our application, as are our basic Classes. If these basic Classes were standardised in the future RDFS recommendation, it would enable all developpers to "wire" them in their code, avoiding possible mismatches and overhead (having to download some other schema, serialize it, and process foo:Integers as if it was another object than bar:Integers). In fact, my understanding is that if many people define the +same+ objects separately and in separate namespaces, it may kill all the potential benefits of namespaces! I feel that the relevant working-group should seriously consider the standardization of a few basic classes, exactly as it has been done (and perhaps for the same reasons) for the possible values of rdfs:necessityValue. The idea that on the long-term, a more general "XML-schema" should solve the issue does not answers to my concern. Best regards Alain Michard INRIA
Received on Wednesday, 5 August 1998 04:51:41 UTC