Just wanted to add some things I forgot in the other mail. I'd like to stress that none of the code you see in the current implementation should be concidered production quality. There are a lot of things that are less than optimal, and a lot of things that are just plain wrong, and might very well blow up. :) Much of it is simply thrown together to test different ideas. One of the things I'm really not sure about, is the character_set_traits concept that is in there now. The basic idea was to allow the library to be used with character sets that are not code point compatible with Unicode by abstracting this into another traits concept, and having the string class use that for it's external interface. This was an idea that seemed good at the time, but I'm getting more and more unsure about the usefulness of it. The biggest reservation I have against it, is that it basically makes it impossible to incorporate Unicode specific functionality in the string class' interface. (Functions for normalization and collation come to mind.) Another thing is the way the Unicode Character Database is implemented. As of now, we simply generate one massive 2MB source file with the database as one gigantic array inside it. This of course leads to equally gigantic executables, which may or may not be desirable. Anyway.. Just wanted to cover my behind before you start complaining. ;) - Erik