Well, for one, it handles text as being UTF-8 by default, even at the lowest level (and in no case assumes that byte = character !)