[Boost-users] Tokenizer design question