- From: <keshlam@us.ibm.com>
- Date: Tue, 6 Jun 2000 10:43:48 -0400
- To: "Tim Berners-Lee" <timbl@w3.org>
- cc: "Jonathan Borden" <jborden@mediaone.net>, xml-uri@w3.org
>I agree that that is a huge problem with XBase. >(Has anyone made that comment formally?) The DOM WG raised a concern about the interaction of XBase and relative namespaces when we were asked to review the XBase proposal. Since the DOM Level 2 design assumed the Literal interpretation, our belief was that XBase should not affect Namespaces. But if validity can be defined in terms of absolutized relative references, as the Absolutize proposal suggests, that means validity is affected by XBase. Note that a relative pointer to the schema (however that pointer is associated with the document) might experience the same effect, so I'm not sure this is new breakage. A relative reference points to a family of resources, with the individual member of the family selected relative to the base URI. If that isn't an acceptable behavior, you shouldn't be using relative in the first place. If some folks still feel it _is_ acceptable -- and required? -- behavior, then we have to decide whether it is actively supported (Absolutize) or passively tolerated (Literal). We still come back to the same basic questions: Do we change namespaces to forbid relative syntax (and declare that a group of previously well-formed documents are now ill-formed), do we change them to absolutize (and declare that a group of previously valid documents are now invalid), or do we drop the requirement to absolutize (and declare that these aren't URI References even though they use URIRef syntax). I can see arguments for all three. I think my personal preference would still be Forbid, Literal, Absolutize in that order, but the space between them has closed somewhat. And I reserve the right to reconsider. ______________________________________ Joe Kesselman / IBM Research
Received on Tuesday, 6 June 2000 10:44:07 UTC