draft-phillips-langtags-08, process, sp ecifications, "stability", and extensions

John Cowan jcowan at reutershealth.com
Fri Jan 7 14:12:29 CET 2005


Michael Everson scripsit:

> >3066bis provides a reliable way to locate country codes in all cases,
> >but the algorithm is different. And this is a non-backwards-compatible
> >change.
> 
> Surely the only thing it HAS to be backward-compatible with is the 
> existing registry.

Ned and others, it seems, have been using the following heuristic:

1)	Country codes are found only in the second subtag of a
	language-tag.

This heuristic, combined with the RFC 3066 rule:

2)	If the second subtag of a language tag is 2 letters long,
	it is a country code.

provides a mechanism for picking out the country code that will not
work in RFC 3066bis, where the rule is:

3)	A 2-letter subtag that is not to the right of a singleton
	subtag is a country code, and nothing else is.

The counterargument is that heuristic 1 only happens to hold in the
existing set of registered RFC 3066 language tags plus those that are
valid without registration; there is nothing in RFC 3066 that compels
it to be true, as tags like mn-Cyrl-RU can be registered by the process.

-- 
"Well, I'm back."  --Sam        John Cowan <jcowan at reutershealth.com>


More information about the Ietf-languages mailing list