University of Hertfordshire

By the same authors

Documents

  • David Castro
  • Raymond Hu
  • Sung-Shik Jongmans
  • Nicholas Ng
  • Nobuko Yoshida
View graph of relations
Original languageEnglish
Article number29
Number of pages30
JournalProceedings of the ACM on Programming Languages
Journal publication date2 Jan 2019
Volume3
IssuePOPL
DOIs
Publication statusPublished - 2 Jan 2019
Event46th ACM SIGPLAN Symposium on Principles of Programming Languages - Hotel Cascais Miragem, Cascais, Portugal
Duration: 16 Jan 201919 Jan 2019
Conference number: 46
https://popl19.sigplan.org/home

Abstract

This paper presents a framework for the static specification and safe programming of message passing protocols where the number and kinds of participants are dynamically instantiated.

We develop the first theory of distributed multiparty session types (MPST) to support parameterised protocols with indexed roles—our framework statically infers the different kinds of participants induced by a protocol definition as
role variants, and produces decoupled endpoint projections of the protocol onto each variant. This enables safe MPST-based programming of the parameterised endpoints in distributed settings: each endpoint can be implemented separately by different programmers, using different techniques (or languages). We prove the decidability of role variant inference and well-formedness checking, and the correctness of projection.

We implement our theory as a toolchain for programming such role-parametric MPST protocols in Go. Our approach is to generate API families of lightweight, protocol- and variant-specific type wrappers for I/O. The APIs ensure a
well-typed Go endpoint program (by native Go type checking) will perform only compliant I/O actions w.r.t. the source protocol. We leverage the abstractions of MPST to support the specification and implementation of Go applications involving multiple channels, possibly over mixed transports (e.g., Go channels, TCP), and channel passing via a unified programming interface. We evaluate the applicability and run-time performance of our generated APIs using microbenchmarks and real-world applications.

Notes

© 2019 Copyright held by the owner/author(s)

ID: 16772736