Why not import await?

why-not-import-await.md

why not import await instead of top-level await?

Some historical discussion

• The proposal
• import await gist
• PR to proposal explainer
• slide from march meeting discussing import await
• Issue for W3C TAG Review
• WebpackPR to add import async

Concerns raised in initial import await gist

• Asynchronicity of imported module is not visible
  • Imports of async modules can have potentially performance impact and should be visible to the developer
  • It's difficult to see which modules are affected by the top-level-await
  • Once this is visible, developer can break the async chain with import()
• Microticks between imports may be unexpected
  • Developer don't know if modules are evaluated sync or async
  • Execution order is no longer guaranteed when modules add top-level-awaits

How the current specification works

• No more microticks unless there is top-level await
  • If no top-level await in graph then it is executed synchronously
• Optmizes async execution
  • Only the modules that include a Top-Level Await are run asyncronously
  • Subsquent sub graphs are run synchronously
  • No micro ticks for remaining graph

Concerns with import await

• Viral Syntax
  • no way to introduce async module into graph without import await to root
  • use of import await defensively likely to happen
• wasm modules will be async
  • will force introduction of wasm modules to be Semver-Major (and viral)
• Removes ability for engine to optimize async execution
  • import await would force graph up to root to execute asynchronously
  • running import async modules synchronously would break developer contract

In Summary

• It is my belief that import await is not a replacement for Top-Level Await
• This does not mean that others, e.g. webpack, should not experiment with the API
• this does not preclude import await from being presented to the committee as an additional proposal in the future