I argue that the characterization of reducibility as computability is too weak to be credible. I propose an improved version -- "intentional computability" -- and show that it is indeed a stronger criteria, in that there may be specifications for which there exist programs but where there is no...