petercon at microsoft.com
Mon May 14 23:38:04 CEST 2007
From: Kent Karlsson [mailto:kent.karlsson14 at comhem.se]
> As it stands, the code would then be 'und', which is the
> only code given for use when the exact appropriate code is
> not known.
IMO, this is the Kent Karlsson view, which reads into ISO 639-2 things that are not there. That standard simply doesn't specify to the level that you're indicating. Extending the logic, one might argue that, if I don't know whether content is 'en US' or 'en GB' or 'en CA' etc. then I'd be forced to declare it as 'und'. Clearly that's not what anybody intends.
My comments here are not meant to be taken as endorsing John's proposal let alone suggesting that ISO 639-2 provides endorsement for his proposal. Rather, I'm simply saying I think it's a bit of a stretch to argue that his proposal is ruled out by ISO 639-2. IETF BCP 47 is an application making use of ISO 639-2 identifiers, not ISO 639-2 itself. It applies some of its own rules, rules that nowhere appear in ISO 639-2. If BCP 47 chose to prohibit the use of 'dgr' for Dogrib and required the use of 'ath' ("Athabaskan languages"), that would be perfectly fine, even though by the terms of 639-2 an individual language code is available and a collective language code is not intended to be used in that case.
More information about the Ietf-languages