Experience with the Larch Prover

Abstract

Many people have argued the importance of mechanical theorem-proving for reasoning about programs. Proving the correctness of programs by hand is usually hard and errorprone. People often miss boundary cases or forget to state hidden assumptions. On the other hand, can current mechanical theorem provers deal with a wide scope of non-trivial problems? Here, the question of scale is in diversity of problems as well as in complexity of each problem. Some provers are more suitable for one class of problems than others and all provers have space and time bounds that set practical limits on the size of an individual problem that can be handled.