So maybe ID isn't a problem after all.

[Tim Bray <tbray@textuality.com>]

So given a couple weeks to think about this, I'm starting
to think that maybe there's no problem.  Given an actual
real XML language [as opposed to an abstraction] you usually
know what the ID attributes are.  It's well-defined for SVG, 
XHTML, and pretty well anything else I can think of.  Clearly 
it's essential that an XLink/XPointer processor have an external
interface by which you can tell it "for this resource, XX is 
an ID attribute".

So where's the problem?  When you're trying to process an
XLink/XPointer into something and the only thing you know
about it is that it's XML.  Er..... what's the scenario 
where this happens?  -Tim