>Quoting from the NS recommendation:
... is somewhat dangerous. There was a _MASSIVE_ W3C debate last summer
about what the intent of the Namespaces spec was, versus what some people
thought it ought to be.
There was no consensus on whether a namespace expressed as a relative URI
should be compared as a literal string, or "absolutized" in the context of
the current base URI. As a result, the use of relative URIs as namespace
names was declared officially Unsupported until such time as this gets
settled -- which probably means waiting for the "Semantic Web" proposal to
actually generate a recommendation, if it ever does.
Unsupported means exactly that: If you use a non-absolute URI name,
_anything_ can happen. It may be taken as a literal. It may be taken
relative to some base URI (it isn't always clear which!). OR, the
application may tell you that you're trying to use an undefined behavior
and should rewrite it in more reliably portable form.
I believe Xerces, and Xalan, have taken that last position. I happen to
agree with them; encouraging folks to write documents which are inherently
nonportable would be a Bad Thing, given that XML is supposed to be about
portable data interchange.
Your milage may vary.
