Browse Books

Reviewer: Jerzy J. A. Klaczak

This is a textbook on algorithms for mutual exclusion, which documents the development of the algorithms on that subject. By no means is it a catalogue or case book, for many of the presented algorithms are of more theoretical than practical interest today. Although it is only a compilation of published algorithms, it will be very valuable to have them all collected in a single book. The book can be used both by teachers and students, since it requires almost no additional background. Chapter 1 introduces us to the fundamental terms and problems: process, competition, cooperation, sharing, communication, protocol, etc. The next four chapters describe in logical sequence the protocols for mutual exclusion: first in the presence of a common memory (both software and hardware solutions), then in its absence (using local state variables and message communication). Real effort was made to present all algorithms in a standardized way. Chapter 6 offers a new perspective, presenting some solutions to the producer-consumer problem and the concurrent reading-while-writing problem. The index adequately lists the names of scientists and algorithms. Nevertheless, it also includes many terms used only incidentally. Still worse, many heavily used terms are associated with a randomly chosen page. One algorithm, the final one of Chapter 5, is only sketched and is not formulated in the Ada-like language otherwise used to uniformly describe algorithms in this book. The author also suddenly (from p. 34 on) changes his style of writing delay loops; likewise he introduces a construct which, although self-evident, was not described in the Foreword. In general, it is hard to find anything in this book, as page headers contain neither current chapter/paragraph numbers nor paragraph names. (The title of the whole book and the chapter name are printed instead.) The style in the chapter concerning hardware solutions in a centralized environment doesn't match the other ones, yet its very presence is required from the point of view of logical completeness. It is also surprising that the problem of fault tolerance was totally omitted in the presentation of just the Ricart and Agrawala algorithm. The most important drawback, however, is the presence of bugs. About two dozen typos—half of them in the program texts—is too much for a book of this size. The book is surprisingly short, considering its thickness and the expectations raised. The book is an English translation of the first French edition [1]. There are, therefore, strange words obviously produced by the translator (mr, ou, bu, etc.); he is probably also responsible for “typical” errors in programs. But there are logical errors in at least three algorithms (pp. 24, p. 31, p. 92–93), evidently made by the original author. Two of them can be easily corrected, and the algorithm on p. 31 will not provide mutual exclusion anyhow. I expect a second edition—which is badly needed—will correct all the mistakes and also include the new algorithms published: (majority) voting, election of a coordinator, etc. It is very hard to recommend this book as a teaching supplement. It completely lacks exercises. Some proofs are not given in full; instead, Raynal has left it to the reader to investigate the Bibliography in order to check the answers. Fortunately the reference list is very exhaustive and most of the entries are widely available.