utf-8 and cultural sensitive sorting

Our product will be storing its character data in utf-8
format (unicode encoding).

What is the best way to achive cultural sensitive sorting
using the utf-8 data?

Is it possible have the locale apply to a connection?

If so, is the cultural sorting support mature in
PostgreSQL?

What type of performance can be expected as compared with
the normal c locale sorting?

Thanks very much,

Steve.