Re: key and document()

Denys Duchier wrote:

> My proposal of the day (:-) is to allow a 3rd optional argument to
> key() to permit specification of a context node (or node set of
> context nodes).  Thus key($name,$key,E) would mean the union of
> evaluating key($name,$key) for each node in the node set denoted by
> expression E.  Naturally, if E is omitted, it would default to
> "current()".

This is a reasonable suggestion. The same thing would work for id() as
well, and is what is hinted at by "an easier way to get an ID in another
document" in the "Features under Consideration for Future Versions of
XSLT" appendix.

James