Another really nice post. If you care about fast _and_ correct computing in C++ (and other relevant languages) but don't follow Daniel -- well, then you're doing it wrong ...