Empirical evaluation of straight-line asynchrony design trade-offs

Build an empirical body of data that supports reasoning about the trade-offs among straight-line asynchrony design choices, including the relative overhead and concurrency of semi-eager versus eager async function application, the bug-prevention effects of simultaneous versus bottom-up cancellation, the degree to which each semantics matches developer expectations, and the relationship between design decisions and programming errors.

Background

The paper develops a nine-dimensional taxonomy of straight-line asynchrony, covering the start and end of asynchronous computations and the behavior of cancellation. Although the taxonomy identifies important trade-offs, the paper does not provide empirical evidence for choosing among competing design points.

The authors explicitly identify several unresolved empirical questions: the runtime overhead and concurrency benefits of semi-eager application relative to eager application; whether simultaneous cancellation prevents realistic bugs that would arise under a bottom-up cancellation strategy; which semantics best correspond to developers’ expectations; and which design decisions lead to more errors. Resolving these questions would help language designers evaluate future async/await designs using evidence rather than informal rationale alone.