- From: Daniel Veillard <daniel@veillard.com>
- Date: Thu, 14 Nov 2002 10:01:35 +0100
- To: "Simon St.Laurent" <simonstl@simonstl.com>
- Cc: www-xml-linking-comments@w3.org
On Wed, Nov 13, 2002 at 04:33:27PM -0500, Simon St.Laurent wrote: [...] > this proposal > inflicts ever more verbosity on XPointers, burdening them with ever more > barriers to human editing and interpretation. > > My favorite example uses the xmlns-local() scheme[1] I proposed > recently. If forced to use this poisonous QName mechanism, what once > looked like: > #xmlns-local()[restOfXPointer] > > suddenly explodes to: > xmlns(ns=http://simonstl.com/ns/bogus)xmlns-local()[restOfXPointer] > > which is amusing at best. > [...] > future participation. I strongly recommend striking this and leaving > such work to future development, wherever it may appear, unless the W3C > can demonstrate that it has genuine commitment to future generic XML > XPointer scheme development which might justify this restriction. > > [1] - http://simonstl.com/ietf/draft-stlaurent-xmlns-local-frag-00.html Heh, good luck Simon. that time I agree with you. The xmlns scheme purpose was to allow interpretation of prefixes when processing XPointers schemes expressions inside document. Reusing them for namespacing schemes mixes layers, and tend to change the way those schemes have to be defined. I still believe that the schemes definitions must be done in tight coupling with the associated Mime Type definition for the resource. Trying to reuse the namespace mechanism to avoid the registry done in practice by the set of RFCs at the IETF is not a practical solution because there is no way to get the semantic or processing description from a random scheme. The key point here is interoperability, and to achieve this I don't see any other mechanism than a well written fragment identifier RFC defining the schemes bounds to a given Mime Type. Proliferation of random scheme definitions without a firm mechanism in place to express their semantic and make sure applications will interoperate in their interpretation of the fragment identifiers (and hence the schemes used) should not be encouraged IMHO. Daniel P.S. While libxml implements the xmlns() scheme, it does not implement that specific new semantic, so it should not used as a proof of implementation for that specific feature of xmlns(). -- Daniel Veillard | Red Hat Network http://redhat.com/products/network/ veillard@redhat.com | libxml Gnome XML XSLT toolkit http://xmlsoft.org/ http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/
Received on Thursday, 14 November 2002 04:23:44 UTC