Browse Books

Reviewer: Teodor Rus

In recent years, concurrent computation and parallel programming became more than a philosophical issue. Theoretical models for concurrent computation—Communicating Sequential Processes (Hoare), Calculus of Communicating Systems (Milner), Data Flow Computing Systems (Dennis), Actors (Hewitt)—were developed, and hardware support for their implementation is provided by multiprocessor computer architectures. While the concepts of sequential computation developed by mathematics so far are based on the principle of substitution, there is no such principle at the basis of concurrent computation. Theoretical models of concurrent computation are expressed in terms of state transitions performed by sequential computations, or partial ordered sets of events. Hence, the concept of sequential computation occurs as a fundamental brick used to develop a theoretical concept of concurrent computation. The actor model uses a partial ordered set of events as a model for concurrent computation. The book discussed here is a contribution towards this idea. Its stated purpose is “. . . limited to theoretical and programming language aspects of the model. . . .” The difference between an actor model of concurrent computation and other models results from changing the fundamental semantic idea of computation from a sequence of state transitions to a finite partial ordered set of events. That is, in an actor system, a computation consists of a finite partial ordered set of events. Each event consists of an actor that behaves according to a script. Since a state transition in a state transition model of concurrent computation can be performed by an actor according to a script, the actor model seems to be powerful enough to generalize other models. Since the semantic of the actor model for concurrent computation is different from the semantic of other models of concurrent computation, the syntax expressing it seems to be different as well. Thus, new names for old meanings are generated. If this is not very visible in the previous research, it appears very strong and sometimes painful in this book. In addition, this book provides no intuitive basis for its non-actor-oriented reader, and thus it is not only difficult to read, but difficult to find a reason for reading. The contents of the book are distributed among eight chapters, as follows: Chapter 1 is devoted to an introduction that reviews the work done in this area at MIT. The big novelty of this research seems to be “talking in terms of processes instead of talking about processes.” However, the supervisor model proposed by Dennis and Van Horn [1] does the same thing. Chapter 2 is devoted to design decisions (read actor language design decisions). These decisions are studied with an open eye on the other languages supporting concurrent computations. This study is rather vague, and the closest languages to this model (Ada and Smalltalk) are completely ignored. Therefore, it is difficult to see the difference (or the relation) between task in Ada and task of an actor, abstraction in Smalltalk and abstraction in actors, etc. Chapter 3 is devoted to computation in actor systems. The lack of precise concepts of task and behavior, as well as the mixture of semantics with syntax, makes it difficult to understand. Chapter 4, entitled A More Expressive Language, helps in this respect. Chapter 5 provides a model for actor systems. Here we can see how mathematics can obscure things instead of illuminating them. Chapter 6 provides a good discussion of the essential issues in a concurrent system. Chapter 7, entitled Abstraction and Compositionality, uses the concept of atomic action to talk about abstractions and provides a concurrent composition law. Chapter 8 contains conclusions. An appendix is enclosed sketching the comparison between actors and the concurrent computation system developed by Milner. Although the title is appealing, promising a new insight on concurrent computation with the actor model, after a few readings I conclude that this book helps you if you know all about it. The language used by the author is poor in both form and content. An example illustrating it follows: Independent systems are connected by sending some buffers for external actors in each module a communication to become forwarding actors which simply send their mail to some receptionist in the module. (p. 107) One can assume that this statement belongs to another section, but statements like it are uniformly distributed among all sections.