A Sound Algorithm for Asynchronous Session Subtyping and its Implementation

TL;DR

This paper introduces a sound, practical algorithm for asynchronous session subtyping that can handle cases beyond previous restrictive methods, supported by an implementation and empirical evaluation.

Contribution

It presents a novel, sound algorithm for asynchronous session subtyping that is not limited by previous restrictions, along with a tool for practical testing.

Findings

The algorithm can handle complex subtyping cases previously unmanageable.

The implementation demonstrates the algorithm's feasibility in real-world scenarios.

Empirical results show acceptable time and space costs for the algorithm.

Abstract

Session types, types for structuring communication between endpoints in distributed systems, are recently being integrated into mainstream programming languages. In practice, a very important notion for dealing with such types is that of subtyping, since it allows for typing larger classes of system, where a program has not precisely the expected behaviour but a similar one. Unfortunately, recent work has shown that subtyping for session types in an asynchronous setting is undecidable. To cope with this negative result, the only approaches we are aware of either restrict the syntax of session types or limit communication (by considering forms of bounded asynchrony). Both approaches are too restrictive in practice, hence we proceed differently by presenting an algorithm for checking subtyping which is sound, but not complete (in some cases it terminates without returning a decisive verdict). The algorithm is based on a tree representation of the coinductive definition of asynchronous subtyping; this tree could be infinite, and the algorithm checks for the presence of finite witnesses of infinite successful subtrees. Furthermore, we provide a tool that implements our algorithm. We use this tool to test our algorithm on many examples that cannot be managed with the previous approaches, and to provide an empirical evaluation of the time and space cost of the algorithm.