Axel Beckert <a...@debian.org> writes: > Hrm, a serious thought on this: Why not implement both variants?
> What if we > * make unknown-locale-code look at ISO 639-1, 639-2, 639-3 and even > 639-5 for generally valid codes, and then > * add a new, maybe pedantic-level warning which is only emitted if a > language group is used in a locale name, i.e. check locales against > ISO 639-5 and if one of these (which IIRC include the language groups > present in ISO 639-2) is used as locale, we emit a tag which might > be named locale-uses-language-group-code or similar? > This currently sounds if it would make use of all our arguments for > and against including ISO 639-2, would be backwards compatible and > more precise and helpful. Oh, this is a great idea. I like this. That way if someone is using a language group on purpose, such as in this case, they can just override or ignore the tag. FYI, the only tags found in 639-2 that are not in 639-3 plus 639-5 are: { "alpha_3": "cnr", "name": "Montenegrin" }, { "alpha_3": "him", "name": "Himachali languages; Western Pahari languages" }, { "alpha_3": "qaa-qtz", "name": "Reserved for local use" }, According to https://iso639-3.sil.org/code/cnr, cnr is in 639-3, so the iso-codes data for it may be out of date. Likewise, him is apparently now in 639-5. The conclusion I'd draw from that is that there's probably no need to add 639-2 if you include both 639-3 and 639-5, and it may be simpler to just ignore it. -- Russ Allbery (r...@debian.org) <https://www.eyrie.org/~eagle/>