This paper is motivated by the fact that the state of a distributed system is determined by the state of each of its components, which can be interdependent. Ensuring that such a distributed state is consistent is very difficult. The paper introduces a fault-tolerant programming model, called transactors. This programming model supports reasoning about loosely coupled distributed components, running in an unreliable environment such as the Internet, from a semantic point of view. In addition, the paper presents examples to illustrate the transactor model. This programming model is defined in terms of the tau-calculus formalism.

Transactors are not concerned with the atomicity, consistency, isolation, and durability (ACID) properties of transactions; the model focuses its attention on ensuring the consistency of the distributed state in the presence of certain types of node and network failures.

This is an interesting paper, because it provides a model to reason semantically about distributed state consistency in distributed systems. Moreover, I believe that the examples in the paper illustrate a possible way to relate the model specification with the verification of the system properties in case of failures. However, a solid background in tau-calculus will be required to completely understand the paper.

Other methods, like Petri nets, input/output automata, and temporal logic, can be used to reason about fault-tolerant distributed systems. However, the transactor model proposed in this paper maintains dependent information needed to detect semantic inconsistencies. I consider this model to be better oriented to represent the operational view of the systems (namely, process creation); it outperforms others by considering the semantics of the system.