There was similar proposal in ietf-822 ML on MIME.
> We are now testing "selecting languages" by the "Accept-Language"
> with http server. I think It can be performed by using mime-type, too.
The proposal used "Content-Language:" header.
> Suppose a person want to read all information in English. If
> default of that pages is written in Japanese, he needs to search
> anchors to the English pages or English version of that pages ( or
> learn reading Japanese :-)).
> Furthermore, I think auto-switching is crucial for autopilot type
> clients(e.g., knowbot).
The problem is how to express language.
There is ISO 639 two letter langage code (not contry code) but it does
not contain a lot of languages.
So, whether it should be extended or not and how can it be extended
is, I think, the largest issue.
Do you want to distinguish Queen's and American English?