Serbo-Croatian continuum: the top level

Gordon P. Hemsley me at
Mon Mar 3 18:58:24 CET 2014

On 03/03/2014 11:03 AM, Doug Ewell wrote:
> Gordon P. Hemsley <me at gphemsley dot org> wrote:
>>> Implementers of BCP 47 who assume all language subtags must be ISO
>>> 639 code elements are just as mistaken as implementers of RFC 1766
>>> and 3066 in the Olden Days who assumed all tags would look like "xx"
>>> or "xx-XX".
>> Mistaken or not, we are still having to deal with the reality of the
>> existence of these Olden Days implementations, which makes it
>> difficult to use script subtags, let alone new primary language
>> subtags greater than length 3.
> As RFC 4646 introduced both the script subtag mechanism and the
> "registered" primary language subtag mechanism seven and a half years
> ago, and as script subtags sort of existed even before then, via whole
> tags registered under RFC 3066, I'm not too sympathetic to the notion
> that we should avoid using the "newer" features of BCP 47 because
> implementers haven't had time to catch up.

It's not that they haven't had the time. It's that they haven't made use 
of it. (A lot of implementers are fairly ignorant about BCP 47, and 
languages in general, despite the importance that localization has in 
their continued success.)

The reality is, many implementations have not been updated past 'xx' and 
'xx-XX'. And, sympathetic or not, you can't ignore reality. Introducing 
a primary language tag of 5 to 8 characters, regardless of whether 
you're strictly "allowed" to, would be extremely disruptive to a lot of 
extant legacy implementations. Extreme caution should be used in taking 
that approach; it is not a decision that should be made lightly.

Gordon P. Hemsley
me at

More information about the Ietf-languages mailing list