- From: David Carlisle <davidc@nag.co.uk>
- Date: Fri, 19 May 2000 10:45:45 +0100 (BST)
- To: timbl@w3.org
- CC: xml-uri@w3.org
> That (full DNS name) is best because it protects against the URI being > interpreteded accidentally on another system. However, in practice > many systems don't bother to or want to know thir own names > and so just use > > file:///users/davidc/file.xml > > When this is being done by an XML processor on a given machine > then so long as it is done consistently it won't matter which is done. > > But that is exactly the point. It "won't matter" what the base URI is for the purposes of constructing relative links, because they are all functionally equivalent. But for namespace use you have just confirmed that with the absolute approach my example has _no_ specified namespace. Different systems are allowed to infer different base URI and so produce different absolute URI for the namsepace. Even under the absolute approach absolute namespace uri are compared as strings not by functional equivalence. So, I can't say what namesapce the elements in the file are in, I can't query the document or style it by namespace aware tools unless I use the rather unreliable process of ensuring that the stylesheet is on the same machine as the document and then use a suitable relative URI to declare the namspace in the stylesheet, then just hope that the XSL system infers the same absolute URI from the relative URI in the document and the relative URI in the stylesheet. But you have confirmed that there is nothing in the uri spec that says I can definitely rely on this working. "absolute" is just "forbid relative URI" in disguise. That would be OK (there is no real need to use relative URI for namespaces) except for the fact that you appear to be _encouraging_ the use of a relative URI to point to a schema. How can that possibly work with the absolute approach? David
Received on Friday, 19 May 2000 05:46:19 UTC