An implementation of the FTAG model in concurrent ML

Non-imperative programming models can simplify the development of fault-tolerant software, in part because of their potential for automatically generating concurrent implementations. This paper describes the design of a concurrent implementation of FTAG, a previously-described functional model for writing fault-tolerant software based on attribute grammars. The implementation involves translating FTAG programs into Concurrent ML (CML), a strongly-typed version of ML supporting multi-threaded execution. The program convertor that realizes this translation is described, together with node and workspace managers that manage the computation tree and stable storage abstraction, respectively.