ok stephen, maybe you can explain this to me. i've done much of the coding towards using unicode internally. but i'm running into problems with all the font-handling stuff in faces.c, which is completely done in terms of charsets. in the unicode world, there will be no leading bytes. i am redoing the places that use an array of fonts indexed by leading byte to use a hash table, but that doesn't solve the underlying problem that fonts are still being determined by doing things like "find all the charsets in a string and then make sure there's a font available for each charset". what should the model be instead? and how to do this in a way that works both with the old X font code and the new xft code? ben