Two agents possess the fishing rights to a lake. Each period they have two options, to catch without restraint, e.g., to use a fine-mazed net, or to catch with some restraint, e.g., to use a wide-mazed net. The use of a fine-mazed net always yields a higher immediate catch than the alternative. The present catches depend on the behavior of the agents in the past. The more often the agents have used the fine- mazed net in the past, the lower the present catches are independent from the type of nets being used. We determine feasible rewards and provide (subgame perfect) equi- libria for the limiting average reward criterion using methods inspired by the repeated-games literature. Our analysis shows that a 'tragedy of the commons' can be averted, as sustainable Pareto-efficient outcomes can be supported by subgame perfect equilibria.