Not that I would rain on anyone's useful little hack, and certainly to be preempted by James Clark is no disgrace. But I would point out that Jing already does this. If it discovers that the schema you feed it is XSD rather than RNG, it tells Xerces to do the Right Thing with it, completely ignoring any xsi:whatever hints.

This fact seems to be little-known, although I called it out at one session or another at XML 2005, in answer to someone who wanted an XSD validator that allows any-against-any validation.

Yes, indeed you did. And I promptly forgot. Oh, well. Maybe my little hack will still be valuable to folks with documents that need to be validated against schemas from several namespaces. And if anyone ever answers the "lax validation" question, I'll add a switch for that, too.