The spec defines a xforms:mustUnderstand attribute that can be placed on _any_
element in _any_ namespace. Argh.
We might need some xtf-like functionality for attributes (possibly also need for
Here's a take on how it could be done: Before sending xforms-ready, we should do a
document.evaluate("//*[@xf:mustUnderstand = 'true']",
And then run through all the returned nodes, and figure out whether they are implemented...
Question is what that means. I wonder if we can ask nsIContent for example whether it has a native implementation or an XBL binding. Maybe just whether it is possible to QI to a given class?