This whole thread just reconfirms my view, stated a couple of weeks ago,
that the current spec is hopelessly informal and we need some PhD student to
sit down and produce a version in Z or something similar.
Incidentally, if we want to be picky (and we do), a corrollory of the first
assertion above is that a parser (read XML processor) is not required to
distingush the data it is passing to the application because the XML-REC
says it must, from data that it's passing to the application because it's
having a bad day.
Mike Kay