Lucas Pardue

@icing can you try this branch out to see if it addresses your failing curl tests?

Thanks for the report Wesley. I wonder if this is another thing that will be fixed by https://github.com/cloudflare/quiche/pull/2080

I think this would cause some issues for the potential future use of NEW_TOKEN frame, so needs some consideration. Possibly just a TODO comment so we dont forget but needs...

I think this is a different manifestation of your prior https://github.com/cloudflare/quiche/issues/1573 issue. Basically, the stream write is probably blocked on send capacity arising from cwnd, because we write HEADERS atomically...

Looks broadly good. It got me wondering if, while we're touching this area, maybe we should do a little more refactor to ensure the time metric is reported on stream...

Right. The duration is cumulative across all HEADERS, while being actively blocked can happen only on an individual HEADERS. For example, 103 early hints, 200 OK, trailers. The audit stats...

I agree, that's why I think it would be better to switch the metric proposed here to a bool that indicates if HEADERS were blocked, or not, at the time...

Since we're talking Cloudflare quiche, I'd like to clarify that example clients and servers (`client`, `server`, `http3-client`, `http3-server`, `quiche-client`, and `quiche-server`) are not designed to maximise performance. So the conclusions...

Yep. As @rmarx EPIQ paper pointed out, implementations are diverse and constantly evolving. It would be super good for reproducibility to list all characteristics of the implementations under test e.g....

There's benefits to both levels of testing. I like the idea of an "iperf for QUIC", I'm not sure how different the proposal is to how we use "HTTP/0.9 over...