LANGUAGE SUBTAG REGISTRATION FORM (R3): pinyin

[Frank Ellermann]

Michael Everson wrote:

>> I'd strip the remark about "omitting Latn", however.
 
> I could live with that but I think in the real world
> people will omit it.

That's precisely the problem, as user you could be 
interested in "any zh-Latn".  Simple right-to-left
matching won't find zh-fonipa.  Hopeless case, we
can ignore it.

A smart matching process could "know" that zh-fonipa
is actually a shorthand for zh-Latn-fonipa.

But it only knows this if the registry contains the
info.  Or if the application has hardcoded knowledge
about this special case.

Counting pinyin we're already at three special cases,
all following one "treat *-x as *-Latn-x" pattern.

There will be more special cases with other matching
patterns, "hardcoding" does not scale.  

> What's wrong with tagging bo-Latn-pinyin?

Nothing.  But bo-pinyin is a problem when looking for
bo-Latn.  Generic variants instead of an extension
also have the strange effect that zh-fonipa-pinyin
is syntactically allowed.  Unlike zh-FR-GB.  With an
extension we could express what makes sense.

 Frank