> I would like to keep the door open. However, I am not sure if nesting of
> grammars does not prevent such future mechanisms. If you convince
> me that the door is open, I can probably be persuaded.
As a person who implements your "divide and validate" framework, I'm
pretty sure that nested <grammar>s are nothing to do with m12n. As James
Clark said, it's just another kind of pattern; the only difference is
that you can name patterns and therefore you can create cyclic graph of
patterns.
Your "high-level" mechanism can be something like
1. introduce "targetNamespace" to top-level <grammar>
2. remove "ns" attribute.
3. add a pattern to switch to the different namespace.
<any />, maybe?
I guess that's it. And as you see, all we need to do is check the value
of "targetNamespace" --- just like RELAX.
--
Kohsuke KAWAGUCHI +1 650 786 0721
Sun Microsystems
