Re: [Ietf-languages] [Ltru] Propose adding ak as prefix to variants akuapem and asante

[Doug Ewell <[e-mail]>]

Martin Hosken wrote on the LTRU list:

> It is common practise (e.g. CLDR) to treat the language properties of
> a macro language as if it were one (the perceived 'largest') of the
> languages in the set. So for example, ak is treated as equivalent to
> tw. Since tw may take the variants aquapem and asante, I propose that
> ak should be able to take them too.
>
> I realise this may be slightly contentious in terms of should one
> treat ak as tw. But since CLDR and various other tagging systems do,
> I see no harm in allowing ak to take the variants.

One of the main benefits of tagging content as to language is to allow easy, reliable, and consistent matching. It’s not normally considered desirable to intentionally create “duplicate” or “alias” tags for the same linguistic entity.

There are exceptions to this principle: extended language subtags (so one can equivalently write “yue” or “zh-yue”), legacy subtags that have been deprecated by the core standard (“iw” or “he”) or by the transition from the old RFC 3066 registry (“i-lux” or “lb”), and perhaps others. But the goal is generally to avoid new exceptions.

For example, people occasionally ask for a region subtag ‘UK’ to serve as a duplicate of ‘GB’, to allow usage such as “en-UK”, on the basis that the country is popularly called “the UK” or that its ccTLD is “.uk” or — fallaciously, with respect to ISO 3166-1 — that the term “Great Britain” does not include Northern Ireland, or some other justification. But the only real purpose this would serve for BCP 47 language tagging would be to complicate matching; tools and libraries would need to add a special exception outside of what is in the Registry.

The Wikipedia article on “IETF language tag” was edited several years ago to claim that script subtags could be considered an “aliasing” mechanism, so that (for example) “yi-Hebr” could be considered an alias of “yi”. I went to some lengths to reword that, because “alias” tends to imply a mechanism meant for convenience, and the whole point of Suppress-Script is to reduce the likelihood of this happening.

“ak/aka” and “tw/twi” are distinguished from each other in all of ISO 639-1, 639-2, and 639-3. In 639-3, Akan is a macrolanguage that encompasses Twi (and also Fanti). The CLDR TC can decide to override this distinction in CLDR and conflate a 639-3 macrolanguage with its largest encompassed language, but I personally don’t think that’s the approach this group should take, as the gatekeepers of the LSR with more or less a commitment to uphold the classification judgments of the core standards.

As some will point out, BCP 47 does not forbid any variant from being used with any prefix one chooses; for example, one could legally write “fr-akuapem”. The Prefix field is a recommendation, not a law. But the suggestion here is to add prefixes that would imply equivalence between the two resultant tags. Most variant subtags with multiple Prefix fields intend that the resultant tags are interpreted differently; “pt-ao1990”, say, is not the same as “gl-ao1990”. This proposal would enshrine this sort of equivalence between Akan and Twi in the LSR.

I’m not the Reviewer, and his decision is the one that matters here, but if it were me, I’d reject this request.

--
Doug Ewell, CC, ALB | Lakewood, CO, US | ewellic.org