Unicode collation is very complex. If you wanted to sort according to that, then you would need to devise an isomorphism between unicode strings and bitstrings such that lexicographic ordering in the bitstring domain agrees with unicode collation in the unicode domain. I'm not saying it's impossible, but it's not at all obvious how to do it. On the other hand, if all you want is some total order for acceleration structures, then you can indeed just use an arbitrary encoding like utf8 and do the obvious thing.