Dr. Andrew Birrell will be giving a talk at the parallel architecture and programming course in room 105 this Monday from 1:30 to 3pm. This talk is open to everyone (see the talk abstract below). Andrew is an early key member of the Xerox Parc CSL where Altos PC was invented. His best known projects and publications include Grapevine (1981, which included the first distributed naming system); his paper with Bruce Nelson about implementing RPC (1983, for which they received the 1994 ACM Software System Award); and his paper on programming with threads (1989). ------------------ Mutual Exclusion: Some History, Some Problems, and a Glimmer of Hope Andrew Birrell and Michael Isard Microsoft Research In common with many system researchers, we have been worrying about how mainstream programmers will be able to write concurrent programs: concurrent programs that work correctly, that take appropriate advantage of multiple processors, and that will continue to work correctly after the superstar programmers move on to their next project. Obviously, the motivation for this comes from the increasing use of interactive applications that use web-based services, and from the advent and prospect of increasingly parallel mainstream computers. Our worry has taken two forms, both of which will be explored in the talk. First, we have looked back at the historical development of concurrent programming, to try to understand the decisions that led us to our present state-of-the-art. Second, we are looking forward to techniques that have promise to correct the present dismal state. We hypothesize that by reducing or eliminating the requirement for programmers to write explicit mutual exclusion code, we can make progress. We present techniques that hold promise in this direction, including a new proposal that we call "Automatic Mutual Exclusion".