Re: Charsets revisited

> > By lowest bs rule server will ALWAYS send iso-8859-5 representation
> > because it about two times less then Unicode one (for russian).
> 
> How is this actually implemented? Would that be that for every
> charset that you have the number of characters in there, and
> then you chose the charset of the acccept-charset  with the best fit?
> Where do you have the database?
> 
> I think this is a bit unpractical, also the heuristics of it.
> Maybe it should be that the list of accept-charsets are ordered by
> preference and the first one that fits is the one transferred.
> There should also be fallback rules, if no exact fit is encountered.

This thread started from my question about lack of quality
factor in recent draft for Accept-Charset header.