Under the assumption of complete rationality, Nash equilibrium is the only reasonable strategy (set) of the finitely repeated prisoner's dilemma. In fact, some strategies only slightly deviate from the so-called rationality, and the corresponding payoff may much better than that of Nash equilibrium. This article points out, even under the rational assumptions, the players have reason to seek a mutually beneficial agreement (Pareto dominated compare to Nash equilibrium) and a weak and optional constraints, so that the agreement can be successfully implemented. If the constraint does not harm the interests of the participants, or the adversely affects of the constraint are negligible, then the finitely repeated prisoner's dilemma becomes a bargaining problem issues on the strategy sequences and the problem to seek the constraints. The quantification of the constraints, the so-called security deposit in this paper, is nearly a concept of distance from an agreement (a strategy set) to the complete rationality.