I have been interested in concurrent programming since about 1963, when its associated problems contributed to the failure of the largest software project that I have managed. When I moved to an academic career in 1968, I hoped that I could find a solution to the problems by my research. Quite quickly I decided to concentrate on coarse-grained concurrency, which does not allow concurrent processes to share main memory. The only interaction between processes is confined to explicit input and output commands. This simplification led eventually to the exploration of the theory of Communicating Sequential Processes.

Since joining Microsoft Research in 1999, I have plucked up courage at last to look at fine-grain concurrency, involving threads which interleave their access to main memory at the fine granularity of single instruction execution. By combining the merits of a number of different theories of concurrency, one can paint a relatively simple picture of a theory for the correct design of concurrent systems. Indeed, pictures are a great help in conveying the basic understanding. This paper presents some on-going directions of research that I have been pursuing with colleagues in Cambridge – both at Microsoft Research and in the University Computing Laboratory.