Designing fault tolerant software with a recovery meta program

Most of the systems intended to support software fault tolerance in application programs (AP$) adopt one of two complementary strategies: (1) a set of low-level facilities (e.g. replicated objects, reliable communication protocols, and reliable file manipulation) is provided to the AP programmer, who has to coordinate their use within his/her software; (2) some particular language construct is provided to take checkpoints, and perform rollbacks. The authors propose an approach which combines both strategies; they describe a programming environment which exploits the difference in the concerns in designing an AP and making it fault tolerant