2.0 TODO

[ronag]

This is a meta list for stuff to look into before 2.0 release

• fix: requestTimeout monitors from sent #343
• body is lacking 'error' handler while queued in pool #356
• fix: CONNECT calls onUpgrade multiple times #348
• signal is lacking 'abort' handler while queued in pool #350
• Review busy and kResuming
• Should onConnect be invoked again when retried due to head of line failure? (current behavior). YES
• Should abort signal and request timeout be moved out from core API and instead explicitly be implemented in stream, pipeline,request. refactor: signal and timeout + remove Client requestTimeout option #349
• 100 coverage
• Move "core" (client, request, errors, utils) into a core folder. (refactor: move core into its own folder #351)
• Is it possible for onBody to return nully?
• Review assert(!this.aborted).
• General review and test improvements.
• Check benchmarks.
• Add test for abort while in Pool queue.
• ...

3.0

• rename onInit => onPending and onConnect => onRunning?
• Consider whether Client "busy" API needs to be changed or improved to allow further pipeline scheduling (i.e. Pool) experimentation in userland.

[ronag]

I don't think I will be able to sort out all potential semver major changes in a timely fashion. Would like to make a release once the bugs (#350, #356) and existing PR's has been sorted out.

The way I see it we have the following things that might be breaking in addition to the existing PR's:

• Client.busy property
• Client.'drain' event
• Client.dispatch method

Would like to leave that to a 3.0 version. Should we mark them experimental in the meantime or is that unnecessary?

@mcollina wdyt

[mcollina]

I'm good with bumping as many semver-major we need :).

[ronag]

Made a 2.0 milestone for remaining issues.