Developing Correct Concurrent Software Using Types

Modern society runs on concurrent software: different processes (threads) jointly process massive data sets and serve many clients and users simultaneously.

Good methods to ensure the correctness of concurrent software are lacking due to the enormous space of concurrent executions.

But it is vital to have some correctness guarantees, e.g., every thread will eventually perform an action (liveness) or private data cannot leak to an attacker (non-interference).

Recent years saw an active development and industry adoption of new programming languages that automatically enforce correctness guarantees through a type system that disables programmers from writing bad programs.

Yet, existing concurrent programming languages cannot enforce deep concurrency properties such as liveness and non-interference.

Moreover, there is no guarantee that the good properties of high-level programs are preserved after compilation to executable machine code, because compilers perform increasingly complicated unverified optimizations.The COCONUT project will design and verify new type systems, compilers and logical abstractions for fully-fledged concurrent programming languages that automatically enforce deep program properties.

COCONUT will address this objective through rigorous machine-checked foundations, leveraging recent innovations in the development of powerful program logics for concurrency (in which I played a key role).

The foundations are essential to prove the validity of our results, but they will also play a key role for experiments and to bring together recent results from the fields of programming language theory, concurrency theory, compilers, security and proof assistants.COCONUT will impact the development of future concurrent programming languages but also the practice of concurrent programming itself, and thereby contribute towards the grand challenge of enabling software engineers to write correct and secure concurrent software.