- From: <keshlam@us.ibm.com>
- Date: Mon, 5 Jun 2000 14:17:01 -0400
- cc: "xml-uri@w3.org" <xml-uri@w3.org>
I've been incommunicado for a week. Catching up on this is going to be timeconsuming. If there's been a large shift in positions/rationalles, I would _GREATLY_ appreciate it if someone could post a summary of the current state of the world, similar to my atttempt to summarize where we were before I left. At that time, I think I was just starting to understand the argument that "the Namespace Name is a URI, the Namespace Declaration is a Reference to that URI, therefore we should absolutize"... but had some serious concerns about whether URIs really met the needs of namespace-aware document processing. * One of those concerns was that there didn't seem to be a strong "is not equal" test for URIs, which Namespaces require. * The reason namespaces switched to references in the first place was to support fragment-identifier suffixes. However, URIs per se don't support these... so if we're talking about the URI being the "real" identity of the Namespace, what happens to that #whatever suffix? (I'm not sure anyone is actually using them... but that's what we thought about relative references. The Namespace spec, as written, permitted them.) * Others had pointed out that some URIs, such as mailto:fred, are "relative" even though they don't use relative syntax. But the "if it hurts when you do that, stop doing that" argument seems valid for that case; if you want a reference to a specific namespace rather than a family of namespaces, you shouldn't use relative syntax and you shouldn't use these. * On a purely "make it affordable" basis, I was concerned that absolutizing meant spending cycles on every namespace in order to make the fringe cases -- the relative references -- work "properly". Even though my employer sellls hardware, I'd prefer to keep costs of processing XML down whenever that can reasonably be done. (I'm a software engineer, not a computer scientist; I'm trained to sensitive to computational overhead even when it it's a small constant multiplier rather than N-to-some-power.) ______________________________________ Joe Kesselman / IBM Research
Received on Monday, 5 June 2000 14:17:14 UTC