I think I'll put in some stronger language.
The content-language tag is NOT intended to help choose the
representation of the characters; information about character
set should ideally be enough to do that, and in cases where
insufficient information is available from the character set definition,
heuristics based on "content-language" are likely to generate
mistakes for some cases.
I'm beginning to agree more and more with Ohta-san here; in my local
locale, there is surely going to be some "##$%&/ reading this and
(implicit) content-type: text/plain; charset=us-ascii
actually means the Norwegian variant of ISO 646, and I DON'T WANT THAT!
(let them use charset=x-iso-646-60 instead!)