Software Synthesis is Hard – and Simple

Abstract

While the components of distributed hardware systems can reasonably be assumed to be synchronised, this is not the case for the components of distributed software systems. This has a strong impact on the class of synthesis problems for which decision procedures exist: While there is a rich family of distributed systems, including pipelines, chains, and rings, for which the realisability and synthesis problem is decidable if the system components are composed synchronously, it is well known that the asynchronous synthesis problem is only decidable for monolithic systems. From a theoretical point of view, this renders distributed software synthesis undecidable, and one is tempted to conclude that synthesis of asynchronous systems, and hence of software, is much harder than the synthesis of synchronous systems. Taking a more practical approach, however, reveals that bounded synthesis, one of the most promising synthesis techniques, can easily be extended to asynchronous systems. This merits the hope that the promising results from bounded synthesis will carry over to asynchronous systems as well.