- From: Mark D. Anderson <mda@discerning.com>
- Date: Sat, 21 Nov 1998 12:25:35 -0800
- To: <rdf-dev@mailbase.ac.uk>, "Tim Bray" <tbray@textuality.com>, <connolly@w3.org>, <www-rdf-comments@w3.org>
(sorry for possible multiple recipients. be careful in replies. follow-ups to rdf-dev only.) In reading http://www.w3.org/TR/WD-rdf-schema , now entering last call, the main question that comes to my mind is: Why? I mean: why do we need this proposal at all? We now have DCD, SOX, XML-Data, XSchema, and RDF Schema, and that is without looking very hard. I could understand this divergence if there were some principled separation of capabilities for different problem domains. UML, IDL, RDBMS metadata, DTDs, and library systems constituent a daunting range of capabilities to trying to bring to XML all at once. But that isn't what is done with these proposals. Rather, they have a seemingly ad hoc admixture of capabilities, such as document inheritance, type inheritance, primitive data types, structural constraints, and so on. Nor should it be surprising that it is hard to find such a convenient partitioning. As someone noted in a white paper i read not long ago, "one person's metadata is another person's data". RDF properties (er, statements) are XML. There will be some generic XML-based alternative to DTDs. Why should RDF make up its own in addition? I could also understand this wealth of proposals if there were contending commercial interests at work (action sheets vs. behavior sheets, etc.). But DCD and RDF are both W3 activities. The overlap between RDF schema and DCD (and between RDF and XLink) is obvious; this has been observed by others as well. Why was RDF not covered in the recent re-org of the w3 XML effort? Nor does it seem to me that the RDF effort needs RDF schema immediately, if their only goal is to supplant PICS and accommodate the Dublin Core. They can just use a DTD. It isn't as if RDF Schema is rich enough to enable an arbitrary authoring tool to read an RDF schema declaration and then provide a decent GUI for creating property instances. Tools for PICS et al will have to be hardcoded anyway, and so the only requirement is to provide a precise spec for humans, not for automatic tools. It would be different if RDF had a large number of waiting applications that need the capacity for machine-readable meta-(meta-)data. But that apparently isn't the case right now -- while it is for the motivating domains of some of the other XML-based schema proposals mentioned above. -mda
Received on Saturday, 21 November 1998 15:33:41 UTC