- From: <keshlam@us.ibm.com>
- Date: Wed, 7 Jun 2000 15:03:35 -0400
- To: David Carlisle <david@dcarlisle.demon.co.uk>
- cc: connolly@w3.org, XML-uri@w3.org
Quoth Dan: > To go further and say "... so let's prohibit it from the syntax" > is to gum up the mechanism with exceptions based on policy. There is a lot to be said for not designing foot-shaped guns unless you really need them. If we don't need relative syntax, and we don't need locator suffixes (DO we really need them?), then we could switch namespace declarations from being URIReferences to URIs, and get that result without exceptions. Quoth David: >You appear to be agreeing (as I do) with Joe Kesselman's quoted >statement that the literal interpretation is workable and gives the >expected behaviour on relative URI. Actually, what I was saying was that I think one can find workable definitions for _any_ of the three behaviors. I had previously had trouble making the Absolutize solution make sense; accepting that an absolutized relative reference really _shouldn't_ be expected to be recognizable as a particular namespace (because relative references point to "a family of resources") got me to the point where I could accept that, in fact, they didn't have to make sense as long as the user gets what they asked for. I still don't think Absolutize is a particularly useful answer, but I now understand what its proponents expect it to do and why. The big questions on the table are still: * whether relative syntax for a namespace declaration is a good idea at all; * if not, whether we should deprecate or otherwise officially discourage it; * if so, what should be done with documents that attempt to use it. -- If continuing to support relative is important to you ("my customers are already using it"), you probably don't want Forbid but can live with the other two. -- If you consider relative more confusing than helpful ("gimme a ./light -- I meant a beer/light"), you probably don't want Absolutize but can live with the other two. -- If you really want the Namespace Declaration to be a URI Reference ("the namespace's identity is a URI, the declaration just points to it") you probably don't want Literal but can live with the other two. If we could eliminate any one of these camps, we could take the intersection of the remainder. I think we're making some progress in understanding exactly what it is that people like and dislike about each proposal. I'm not sure we've actually gotten to the point where folks are withdrawing their objections... or at least that there's any clear trend therein.
Received on Wednesday, 7 June 2000 15:03:50 UTC