This attribute discussion is interesting, because I'm not sure you want to
expect the processor to have the entire document available for stack building.
One of the projects that I'm working on (slowly) is an implementation of
XPointers that uses the server's processing power to chop just out the section
of a document the user wants and avoiding sending out millions of bytes of
wasted data. I suspect that we'll need to make the server handle the xml:lang
and xml:space stack, and re-enter their values into the topmost element
returned, or something like that.
xml:lang and xml:space seem like they're worth the extra effort, though.
Simon St.Laurent
Dynamic HTML: A Primer / XML: A Primer / Cookies