The study of reliable communications is important not only in the traditional areas of computer networking, but also with the new distributed-memory multiprocessors or multicomputers. These multicomputer systems, as they increase in size and complexity, will experience lower reliabilities. Furthermore, as the systems grow, the opportunity for centralized control, both in operation and in error detection, is limited. Thus we shall soon see an even greater resurgence in the study of reliable communications mechanisms.

To bring new practitioners into the field, a good introductory survey text is necessary. Such a survey should describe the basics of reliable communication and point out research directions. To fill this role appears to be the mission of this chapter.

The chapter provides a fairly sparse discussion of the well-known layered model concept of communication, which leads into an overview of points that need to be taken into consideration when adding reliability to these layers. The problem of “residual reliability,” reliability built on top of reliability, is discussed. This is an interesting practical question and the text provides a basis for some thought in this area.

Perhaps the most significant contribution of this work is a single paragraph that describes recovery in the event of an error. Its point is one which is often overlooked; put succinctly, it merely states that a simple retry of a failed operation will often succeed, due to the nondeterminism present in a computer network. This is often borne out in practice.

The text has some flaws that make it inappropriate for beginners. The figures and diagrams, for the most part, are incomprehensible. Moreover, some of the terminology is nonstandard and some key terms (or buzzwords), such as “distributed control” and “fault-latency,” are missing. The references are old and insufficient in number to make this chapter usable as a survey work.