W3C home > Mailing lists > Public > www-international@w3.org > January to March 2013

RE: Language ranges with more than two sub-tag

From: Phillips, Addison <addison@lab126.com>
Date: Fri, 1 Mar 2013 10:03:36 -0800
To: Marcos Caceres <w3c@marcosc.com>
CC: "www-international@w3.org" <www-international@w3.org>
Message-ID: <131F80DEA635F044946897AFDA9AC34773AB02C6CE@EX-SEA31-D.ant.amazon.com>
> 
> I don't know if anyone here can help me, but what I'd really like to find is data
> that shows what Accept-Language: values are being transferred over the wire. I
> know that this will not be completely representative [x]. But, if I can show that
> at least some people are, by default or not, being excluded, it could weight
> heavily towards swaying browser vendors.
> 
> [x] http://www.w3.org/International/questions/qa-accept-lang-locales


This article was written in *2003*, which is *before* the current BCP 47 came into use (2006). I'm appalled to find it still out there and I'm sure we'll revise it presently ;-).

If you want data indicating that "some users send three (or more) subtag language tags", all you need to do is open up Internet Explorer's options panel to language and scroll through the list. IE has generated A-L with three-subtag tags for a while now (not the Chinese ones, but quite a list of other languages).

Also, you should note that several locale systems infer additional subtags (particularly the Chinese ones) when they are not provided. 

Various browser vendors are slower to adopt additional language subtag combinations. However, a consistent set of locale identifiers and locale identification rules is called for here. A fallback system that just blithely skips over information that users have provided in their language ranges (particularly script information!) is profoundly unhelpful. Implementing the BCP 47 Lookup algorithm is only a minor elaboration on that.

However, I'll also point out that JavaScript's Intl extension also allows for another model for matching a language priority list (which is what Accept-Language is) that is considered an improvement by most implementers who have worked with it and that's what I would recommend for Sysapps.

(In another message on this thread you said:)

> 2) there are two implementations of the JSON i18n model (Google packaged
> web apps and Mozilla's packaged web apps), so it's kinda already a de facto
> standard. The model used by Google and Mozilla is what the SysApps WG is
> trying to standardise on (hopefully without breaking existing content).

There is an implementation of Widget spec that I am aware of: Amazon Kindle/Kindle Fire uses it. Although it is not a perfect model, it would be painful to change it.

> 
> Agreed. It would certainly make sense to align where possible. However, I'll
> need to ask Norbert for guidance on this, as I haven't fully groked [3] yet.
> 
A better starting point for me to have given would probably have been:

   http://norbertlindenberg.com/2012/12/ecmascript-internationalization-api/index.html


... where Norbert has put some text in English that is more grokable than the specification.

Addison

Received on Friday, 1 March 2013 18:04:13 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Friday, 1 March 2013 18:04:14 GMT