Communicating parallel processes
Jon Kerridge, Dan Simpson · Software Practice and Experience · 1986
Abstract By considering the problem of an event timer it is shown that the commonly available synchronizing facilities (monitors, CSP, distributed processes) are not able to always satisfactorily model the requirements of several processes which must run in parallel and which have to communicate with each other. The problem is discussed in general terms which show that what is required are new concepts for communicating processes. The synchronization facilities proposed are augmented to incorporate the concept of process scheduling directly from a process. This ensures that proper scheduling of process components can take place. The new mechanism is then applied to a number of the standard problems. It is also shown that the use of nondeterminacy in current facilities is probably not required and is, in fact, for many applications, a positive disadvantage.