Rethinking the bounded mpsc channel

[danburkert]

Hi all,

I'd like to bring rust-lang/futures-rs#800 to the attention of the WG. I won't reproduce the entire content of the issue here, but the short version is that the current bounded mpsc channel in futures-rs is difficult to use while preserving back-pressure.

That issue left off with a PoC channel implementation that does a better job at enforcing back-pressure, and shouldn't have much of an effect otherwise on the API or performance. The hope is that proof of concept can serve as the basis or inspiration for overhauling the current version in futures-rs.

Hopefully this is in the purview of the WG, because I think it's a critical component for building robust async systems!

[aturon]

This is for sure within scope! In general I don't think our backpressure story is fully baked.

(The futures team has been primarily focused on the 0.3 work of getting the core Future and Stream traits ready for async/await, but this can absolutely proceed in parallel)

[cramertj]

I'm actively working on a PR for fixing this issue. I'll post here with updates when I have a prototype or a PR ready!

[aturon]

There's some very interesting discussion happening on the PR.