Actor model later history

In computer science, the Actor model, first published in 1973 (Hewitt et al. 1973), is a mathematical model of concurrent computation. This article reports on the later history of the Actor model in which major themes were investigation of the basic power of the model, study of issues of compositionality, development of architectures, and application to Open systems. It is the follow on article to Actor model middle history which reports on the initial implementations, initial applications, and development of the first proof theory and denotational model.

Investigations began into the basic power of the Actor model. Carl Hewitt [1985] argued that because of the use of Arbiters that the Actor model was more powerful than logic programming (see indeterminacy in concurrent computation).

A family of Prolog-like concurrent message passing systems using unification of shared variables and data structure streams for messages were developed by Keith Clark, Hervé Gallaire, Steve Gregory, Vijay Saraswat, Udi Shapiro, Kazunori Ueda, etc. Some of these authors made claims that these systems were based on mathematical logic. However, like the Actor model, the Prolog-like concurrent systems were based on message passing and consequently were subject to indeterminacy in the ordering of messages in streams that was similar to the indeterminacy in arrival ordering of messages sent to Actors. Consequently Carl Hewitt and Gul Agha [1991] concluded that the Prolog-like concurrent systems were neither deductive nor logical. They were not deductive because computational steps did not follow deductively from their predecessors and they were not logical because no system of mathematical logic was capable of deriving the facts of subsequent computational situations from their predecessors

Compositionality concerns composing systems from subsystems. Issues of compositionality had proven to be serious limitations for previous theories of computation including the lambda calculus and Petri nets. E.g., two lambda expressions are not a lambda expression and two Petri nets are not a Petri net and cannot influence each other.

...
Wikipedia