Prisoner's dilemma played by finite automata is reviewed again using a slightly modified measure of complexity. At a first step, an equilibrium with a large number of possible outcomes is shown to hold. At a second stage, we consider a game of repeated interaction, and show that on (limit)...