We address the problem of learning and implementation on the Internet. When agents play repeated games in distributed environments like the Internet, they have very limited {\em a priori} information about the other players and the payoff matrix, and the play can be highly asynchronous....