* Stephen Colebourne: > PS. How bad would it really be to change NaN == NaN to be true? > Wouldn't any breakages be pretty minor? There's more than one NaN, so it would still be magical. x != x for checking for NaN is also quite widespread. Thanks, Florian