Sort of random question: how difficult is it to write a decent Prolog interpreter? I read somewhere that you need to implement the WAM to get efficiency, but I don't know if that's true, and if it is true, how difficult is it to write a WAM?