> That would have meant that "foo" would match PIs with target
> foo. And "*" is a name test too, and we don't want * to
> match PIs, comments, text nodes and document nodes. But no
> doubt some other way of avoiding those problems could be thought up.
Fair enough.
"All elements, attributes, and namespaces whose name matches" would have
worked, I think, given that it's only for the self and ancestor-or-self axes
that you can actually mix those three. (But of course in 1.0 we didn't have
the ::element() test to match element nodes only, so * was a bit
overloaded.)
Michael Kay
http://www.saxonica.com/
