The problem is solved - half of it, that is. UTF-8 has been accepted for the 0.7 proposal. Any language can be described by it. Now you just need a client that fully implements this proposal, and a computer with your desired default system locale, or a smart client that accepts any character in the UNICODE range.

Just returning specific language servers for a specific language query isn't very desirable and anyway impossible to implement: You would need to explicitly specify the language of a file. This is bothersome, unnecessarily restrictive (think of a word which is written with the same Chinese characters in Korean, Chinese and Japanese) and much too complicated to implement. Also, what would you do with servers which offer mixed language files?

The problem is solved, with the 0.7 proposal. Now it depends on the clients.