Jacek Sieka

Yamux is probably a dead end in general - ie quic does everything yamux purported to do and I don't believe we have any strict requirements on supporting it, once...

> Is there any rationale of having the timeouts in the first place? to avoid too many concurrently open streams - it's a dubious mechanism at best to have in...

> passive sampling happy to when we get to that point in our implementation, though this PR is useful independently (given how TTFB is not actually used correctly today)

> Does it make sense to increase the number of concurrent requests to better utilize 'long fat' (high bandwidth + high latency) connections? eh, this is always a tricky one...

> @arnetheduck Why not just allow a single stream per protocol at a time? It seems easier to implement. right now, there is no limit so all clients implicitly already...

this would require a 5.0 release ..

actually, I thought a bit more about this - since the symbols were deprecated before the 4.x release, we can remove them any time

you will broadly be better off with `close` that does not raise exceptions - libp2p however has to deal with three forms of closing: "half-closing" which usually is known as...

> create utilities that deal with common boilerplate my recommendation is to avoid too many of these - they _tend to_ hide an underlying design problem - ie if you...

> unless we still want to support it. hm, unless there's a strong reason not to, it does look .. convenient to support - ie the resolution _can_ be inferred...