This work considers a sequential program, subject to failures, in which checkpoints may be installed in each of the well-defined program steps at given cost. Recovery, in case of failure, is made with respect to the closest preceding checkpoint at a cost which only depends on the checkpoint chosen. A dynamic programming formulation leads to an algorithm which allows a choice of checkpoints, minimizing the total average execution time of the program. This is an interesting and elegantly done piece of work. It is not really applicable to previously studied cases related to database recovery because of the types of cost functions which the authors have chosen.