A Theory of May Testing for Actors

Abstract

The Actor model and 71"-calculus have served as the basis of a large body of research on concurrency. We represent the Actor model as a typed asynchronous 71"-calculus, called A1r. The type system imposes a certain discipline on the use of names to capture actor properties such as uniqueness and persistence. We investigate the notion of may testing in A1r and give a trace based characterization of it. Such a characterization simplifies reasoning about actor configurations as it does not involve quantification over all environments. We compare our characterization with that of asynchronous 71"-calculus, and highlight the differences that arise due to actor properties.

Verification of asynchronous systems with an unspecified component

“…The take information is removed from the labels. The send actions as the consequence of a message taken by M c are renamed to depict M receiving such messages with the application of the renaming operator ρ c2 where the renaming function c 2 : Act T × Act * s → Act r ∪ Act * s is defined as: Our approach for establishing the LTS characterization of an actor in terms of its interactions with an environment is consistent with the one proposed for π-calculus specifications embedding the actor model [69], where channel names indicate actor names. This approach is based on the concept of Interfaces that maintain actor names to/by which a message can be sent/received, called external actors and receptionists, respectively.…”

“…We would have got a considerable practical gain if such reorderings could had been handled semantically through an appropriate notion of equivalence such as may-testing [69,15] instead of explicit shuffling of the messages in each message body of InfM . Due to the non-existence of a practical tool to compare the traces regardless of their action orders in our experiments, we were confined to follow this approach.…”

“…Both trace and may-testing equivalences [28] are appropriate for reasoning about safety properties [15]. However, a trace-based characterization of may-testing such as [15,69] alleviates reasoning, as it does not involve quantification over observing contexts required by may-testing. Such a characterization captures the essence of asynchrony by permitting receive and send actions to be delayed through the following rules, adapted to our setting:…”

“…To check a given actor M against the generated assumption, we propose an approach to derive its LTS characterization in terms of its interactions with an environment in the same line of [69]. Finally we investigate if the LTS conforms to the generated assumption according to the trace inclusion relation [16].…”

“…To check a given actor M against the generated assumption, we propose an approach to derive its LTS characterization in terms of its interactions with an environment in the same line of [69]. Fig.…”