Kaihsu Tai wrote in part:
Using the hyphenated language tags assigned by IANA, such as zh-min-nan, would conflict with this scheme. For example if we used zh-yue, it would be difficult to know what "yue" refers to. Is it an SIL code or an assigned code?
(option c) No, it would not. RFC 3066 adopts all ISO 639-1 language codes. Just look up in RFC 3066 (and transitively ISO 639-1).
I mean it's ambiguous if you use both the ISO-SIL hybrid and RFC 3066 on the same site. I'm quite aware of the fact that RFC 3066 adopts all ISO 639-1 codes.
We could use the RFC 3066 codes instead. This is still an option. However it wouldn't give us access to a large number of languages without resorting to awkward constructions such as x-sil-RUP.
(option c, corollary) And that is the thing one should do: using constructions such as x-sil-RUP when nothing more appropriate comes up elsewhere in RFC 3066. That is why such standards are useful -- they are usually very precise and expressive, however awkward one may consider.
If I'm wrong about that, feel free to explain it to me.
I have been arguing for the hardcore option (c), but I think a compromise is to consider option (a), which will involve (option a, implementation)
- changing 'minnan' and 'zh-cfr' to simply 'cfr'
- changing 'roa-rup' to 'rup'
[3) any other adjustments]
I suggested a few similar schemes to Brion, and he turned them down on the basis of lack of standards compliance. SIL codes may conflict with future ISO 639 assignments.
[By the way, much of the SIL-inclusion problems have been argued about when RFC 3066 was being drafted ... and the result was x-SIL-RUP etc. Sorry to be speaking a bit ex cathedra, but I resent fighting battles already-won....]
This is a different battle, we have different needs. I'm disappointed you haven't commented on my idea of using descriptive names rather than codes. Cryptic hyphenated codes may be useful for information interchange between computers, but humans often find them hard to remember. We can always include the ISO 3066 code in the document headers, and we can always set up redirects from various language identifiers to the primary subdomain.
-- Tim Starling