Benjamin Saunders

> We did PR this a while ago and the discussion lead to not including it. for our usecase we want to implement multipath and then won't need this anymore....

I'm not aware of a use case for replacing the congestion control algorithm in-situ. If you don't have a compelling use case either, maybe close the issue? If someone has...

We should mark `push_next` as `unsafe`. Perhaps we can justify doing so in a patch release: - it's fixing a serious bug - it's rare to be building Vulkan structs...

I concur that the use case for pushing a whole chain at once is unclear, and a simpler method is appealing. I'm not even sure we should it's worth preserving...

In https://github.com/facebook/buck2/pull/922, I have a unit test which is trying to validate that a future does *not* issue a HTTP request by wrapping it in `tokio::time::timeout` and asserting that it...

The mechanism isn't intended to work with external I/O. When both client and server are hosted on the same runtime, then auto-advance won't be triggered while the request is in...

Actually, let's get https://github.com/quinn-rs/quinn/pull/1962 in first.

Triage: This is still desirable, and is nearly ready. I've gone through and resolved threads which have been addressed. Remaining concerns: - enabling `xcb` by default on Unix platforms may...

Would throttling outbound requests (e.g. #922) be a more graceful solution here? It's interesting to note that cargo itself is routinely used to download hundreds of crates and doesn't seem...

Digging around some, it looks like the root issue here is that the `hyper` HTTP/2 client optimistically (legally) assumes that crates.io will accept unlimited concurrent requests, while the actual limit...