Re: Relative Namespaces

[Jeff Rafter <jeffrafter@earthlink.net>]

Thanks for the comments...

I did some hunting for that quote and I think the issue for me is suddenly
more clouded.  Tim Bray commented on the issue when Namespaces wrt SAX2 was
being discussed [1] in the January thread "SAX2: Namespace Processing and
NSUtils helper class"

[1] http://lists.xml.org/archives/xml-dev/200001/msg00119.html

'Pardon for flogging this possibly-dead horse, but you have to read the
text of the namespace spec carefully to realise that you just can't ever
have a namespace URI whose value is "".  We should have put in a sentence
in section 5.2 saying "Note that as a consequence of this rule, it is not
possible to have a namespace whose value is the empty string."'

The mail cited is pretty much the final word in that discussion regarding
the null namespace issue.  The DOM's interpretation is mentioned a couple of
mails later, interestingly.  This seems totally at odds with the DOM's
interpretation though the DOM wording is loose-- opting out because it "does
no lexical analysis"-- but it seems like it could check for an empty string
as easily as checking for null.

I think I am beginning to see some of the big picture here, but can someone
explain how the Schema view fits in?

Also to clarify-- is Tim saying that the following is pointless because it
equates to a prefixed name being treated as a non namespace name or is he
saying that the document is in error wrt to namespaces because the prefix
can not logically be resolved?

  <foo:bar xmlns:foo="http://foo.com">
    <foo:baz xmlns:foo=""/>
  </foo:bar>

Thanks for all of the patience...

Jeff