The mutual exclusion problem has provided a rich field of research for computer scientists. The problem is easy to state, has real-world application in operating systems, and generalizes in a variety of interesting ways depending on the interprocess communication medium and the degree of exclusion desired. This paper explores resource allocation in the case where k units of a resource are available, a process may request up to h units of the resource, and communication is done via message passing. The authors begin with a brief survey of the standard mutual exclusion problem using message passing, move on to show why the standard solution is inadequate for the new problem, and conclude with their solution to the problem and some performance-related results.

I found this paper unsatisfying for several reasons. First, the authors use a straw man to justify their solution. They show that if the resources are requested one at a time, deadlock may occur using the standard k-mutual exclusion algorithm, but their own solution has processes requesting all of their resources at once. Since the standard algorithm can be generalized to handle the case where all of the resources are requested at once, at least for some combinations of values of h and k, the authors are stacking the deck.

Second, the authors’ solution seems too obvious, like something a graduate student would come up with after 20 minutes of thought about the problem. Sometimes the obvious solution turns out to be the best, and I certainly do not want authors making their results unnecessarily complicated, but a simple answer to a problem demands more explanation, for example a proof that the solution is in some sense the best possible. The authors provide no such explanation, and I cannot help feeling that this is a preliminary report that might have benefited from a little more development. What is in the paper is pretty good, but I wish it had more substance.