2013/5/6 matz (Yukihiro Matsumoto) <matz / ruby-lang.org>:
>
> Everything that can be resolved without locale/language information (for most of the cases).
> Case conversion may have problems with some characters, e.g. german eszett or turkish i, but can be converted mostly.
> Of course it would take time to implement everything, but the basic principle (and goal) will be.

Be careful about network libraries.

Text based network protocol itself is not Unicode based in general.
If a primitive used in a such library is extended to Unicode, it may
cause mismatch to the protocol.

For example, the result of "grep -r strip lib/net" should be examined to
strip is used properly or not.
-- 
Tanaka Akira