This is actually my primary concern with the current namespace proposal: that
the inheritance model used here makes it difficult to extract subsections of
documents reliably. I would strongly prefer NOT to have the client need to
download the entire document containing the subsection referred to by the
XPointer - ideally, the server could just send the piece requested.
(Bandwidth still isn't free, you know.) This model opens up an awful lot of
possibilities, but I fear namespaces may snuff it out.
So to make a this kind of request involving a namespace, the client will need
to expand the prefix within the (foo:bar) which yields URLs like
http://...#descendant(17,http://www.simonstl.com/schema/v1:bar)
Maybe this will work, but it sure makes for a mess. Then the server needs to
keep a namespace-expanded version of the documents it's storing if it wants to
process this with any degree of speed, since parsing the whole document and
expanding namespaces gets to be a chore when you're answering lots of requests
simultaneously.
What I'd hoped to do with a simple Java servlet now looks like an _incredible_
mess. Maybe the XPointer folks have some better ideas...
Simon St.Laurent
Dynamic HTML: A Primer / XML: A Primer / Cookies