ID for language-invariant strings

[Frank Ellermann]

Peter Constable wrote:

> we should not fail to determine requirements and design
> appropriately just because parts of what engineering has
> been done up to now wasn't done so using good engineering
> practice.

It is too late to come up with requirements for RFC 3066,
we can't get rid of the region subtags, and we can't add a
MUST NOT against generic variants.  Some damage control is
still possible.

> zxx versus und. Doesn't that suggest to anybody else here
> that just maybe those people don't agree because this is
> an instance of inappropriate overloading?

Yes, there should be only one tag "zxx", no "und", or v.v.

 Frank