Peter Constable petercon at microsoft.com
Wed Sep 3 20:57:36 CEST 2008

> From: John Cowan [mailto:cowan at ccil.org]

> > I think we all agree that Latin is implied. Chinese is also implied.
> By
> > this rationale, a complete tag of "wadegile" would work just as
> > well as "zh-wadegile" (BCP47 syntax requirements aside). In terms of
> > semantic representation, that is true: "wadegile" contains just as
> > much information as does "zh-wadegile".
> However, all tags MUST have a language subtag, so this analogy is not
> on all fours.

Even if the language subtag weren't required, I think we'd still recommend it be included for reasons I described.

> > Those are made more complicated if "Latn" is not part of the prefix
> for
> > "wadegile" and "pinyin".
> You still need to know that "wadegile" and "pinyin" imply Latin,
> because
> the Prefix for a subtag is only a SHOULD, so people are still free to
> send you "zh-wadegile" whether the Prefix says "zh" or "zh-Latn".

No, I (or some process) does not *need* to know. For analogy, that's like saying that Unicode implementations need to be able to interpret and process U+A866 correctly, which clearly is not what that standard requires. No more does BCP47 require that implementations must know that "wadegile" and "pinyin" imply Latin. And, in fact, I suspect that many *will not* know that, while they would be able to do the expected matching when comparing "zh-Latn" with "zh-Latn-wadegile" or "zh-Latn-pinyin". If we want people to get the best results -- or, as Niall said, the most fault-tolerant behaviour -- then we should be giving them a recommendation to tag "zh-Latn-wadegile" and "zh-Latn-pinyin". The prefix field will be treated as exactly that: a recommendation.


More information about the Ietf-languages mailing list