- From: <keshlam@us.ibm.com>
- Date: Fri, 23 Jun 2000 14:12:00 -0400
- To: "David G. Durand" <david@dynamicDiagrams.com>
- cc: xml-uri@w3.org
>I'm not 100% sure that this is a good idea. My objection is based on >a separate set of issues: > 1. Xlink should work for any well-formed XML document. > 2. XLink is defined on the information set of an XML document. > 3. Therefore, every well-formed XML document should have an information set. Personal opinion, which may not have anything to do with the opinion of the Infoset group: We already have documents which are XML-well-formed but not NS-well-formed (due to NS-mediated attribute clashes). Arguably these do not have valid infosets when interpreted by a namespace-aware processor. I'm not convinced that forbidding relative syntax in the NS declaration is a signficant step beyond that point. If you want to process something without testing whether it's NS-well-formed, treat it as a non-NS document. If you want to ask about namespaces, you need properly defined namespaces. >Q: what does the DOM spec return for the namespaceURI attribute? >A: unspecified; My current inclination is that the NS declaration attibute returns the declaration as written. I think we need that info in the Absolutize model, in order to be able to synthesize the correct prefix when an absolute and a relative NS declaration are both in scope. And in the Literal or Forbid models, as-written _is_ the NS's name, so that's OK. ".... drop back 40 yards and punt...." ______________________________________ Joe Kesselman / IBM Research
Received on Friday, 23 June 2000 14:12:19 UTC