- From: David Carlisle <davidc@nag.co.uk>
- Date: Tue, 16 May 2000 19:31:07 +0100 (BST)
- To: xml-uri@w3.org
Are there any use cases for namespaces where relative URIs are actually necessary? Never necessary, the namespace name is just a string so it doesn't matter whether it's a relative URI or not, although there doesn't appear to be any real harm in allowing them. Jonathan Robie's "playing around" scenario [1] is thought-provoking, but not very convincing, at least for me. Also, are there use cases today where people have really gone ahead with relative URIs namespaces? Yes, Microsoft tools have ensured there's a few thousands or tens of thousands of such documents (whether or not these documents are created for good or bad reasons, one can't deny they exist) Other than fulfilling a dream of consistency, I can't see any good reason to support relative URIs in namespaces - ever. Certainly if the namespace rec was changed to imply an `absolute uri' rather than a string literal interpretation, then relative URIs in namespaces would be essentially useless, even if allowed. I can think of no possible use of a system that means if you move a document from one directory to another, all stylesheets, queries, and processing of that document break. Personally I use local URI sometimes when I want to stress that the elements are not in any globally known namespace, for example XSL stylesheets can contain data stored in elements from any namespace that is not XSL. Being not XSL is the _only_ property that matters here. I have been known to use "/dev/null" as the namespace name. (actually this particular use would work even under an `absolute' as the elements are only ever referenced by the stylesheet, so the references would be made absolute in the same way) If this were banned, I could use something else, so as you say above this is not necessary. But the only reason I've seen for banning relative URI is that they don't have the namespace spec stated goals of uniqueness, but changing /dev/null to file:///dev/null or mailto:davidc doesn't ensure uniqueness either. There are all kinds of namespace URI that may be bad for some (or even most) purposes relative URI are just one of them and they do very little harm especially compared to the harm that a change to the namepsace spec would make. (xpath implies relative namespace uri are made absolute, but it appears that most implementations don't do that, so a change to xpath wouldn't be so serious) David
Received on Tuesday, 16 May 2000 14:31:39 UTC