- From: David Carlisle <david@dcarlisle.demon.co.uk>
- Date: Tue, 13 Jun 2000 00:31:54 +0100 (BST)
- To: james.anderson@mecomnet.de
- CC: xml-uri@w3.org
me> There are two questions, whether there has to be a retrievable me> entity to which I think everyone agrees the answer is no. me> Even if a URI scheme that allows retrieval is used. No, there are those who don't agree with this. crikey, I knew Dan Connolly said it was a bad idea to do that but I didn't know anyone would try to claim it shouldn't be allowed. By specifying a scheme which is associated with a retrieval mechanism, the expressed intent is to identify a retrievable resource. Er no. Strange as it may seem, the expressed intent, when specifying a namespace name, is to name a namespace with that name. In which cases I would judge the expression to be erroneous if the retrieval were to fail. Fortunately no existing namespace aware software asks you to judge. Firstly retrieval will never fail as namespace processing never attempts retrieval (later processing might, and then of course you would get failure, but we are discussing namespace processing). Currently I would estimate that the most common form of namespace name that is actually recognised and acted on by a namespace aware processor is an absolute URI in the http scheme that refers to a non existing file. For example the java implementations of xslt allow such namespace names with the last path component being any class name on your class path, since these class names don't relate to files on the server that is "used" for the extension namespace URIs they all end up pointing at nothing. There are of course far more of these extension namespaces than fixed namespaces like xslt, mathml, xhtml, ... David
Received on Monday, 12 June 2000 19:30:03 UTC