Skip localized versions and go for full UNICODE! This is a world wide web after all. I'm located in Germany but I want files in Japanese. Understandably I normally don't get access to a Japanese language OS, and if I do then I get problems with non-UNICODE aware programs with German characters.

If you do localized versions using the country specific encodings, how would you solve the compatibility problems when going beyond one country border? You need UNICODE, it doesn't work otherwise.

Java has an internationalized version of the runtime environment, which is able to e.g. display Japanese characters on a German Windows (settings have to be modified, though). Java itself is UNICODE based, but still there might be some methods that request a string in the system default locale from the hosting OS. If that happens, then for example Japanese will be turned to ?'s on an English OS. This might be a problem, but Java itself is not.

If you use UTF-8 as the UNICODE transformation format for the search traffic then the protocol will stay compatible with current versions. Old clients will see funny characters, but that they do even now. New clients would recognize UTF-8 and read it as UNICODE.

Please do skip versions with country specific encodings! It would make it nearly impossible to go for a world wide reach later!