bug 258246: RFC: Accept language header: 'en-gb' and matching

[Gervase Markham]

Tobias Burnus wrote:
> Well, in principle our algorithm is according to the spec and therefore 
> already the best.

Interesting definition of "best" :-) Most web browser implementations de 
facto partially disagree with this...

>> Whatever, I think that if someone sends de-de and we don't have a 
>> de-de but do have a de, we should certainly match that rather than 
>> failing or falling back to the default. Makes perfect sense.
> 
> i.e. de => ./de-*/, de-de => ./de/ and (of cause) de-de => ./de-de/ and 
> de => ./de/ ?

If you get de-de, you first try de-de, then de, then fall back.

> Well, then there are also the long and the short signs 
> (traditional/simplified), I frankly don't know in how far they pose 
> problems, but I'd expect that most Chinese prefer Chinese to English 
> independed of the kind of signs.

I don't think we want to start matching a sent zh-TW against zh-*, e.g. 
zh-CN. That way does like problems, I agree. But falling de-de back to 
de (i.e., matching the stem exactly, rather than by prefix) seems 
reasonable to me.

Gerv