Max Inden

I am hesitant exposing the internals of `Histogram` just for the use-case of testing. Unless we come up with better use-cases, I don't think we should merge here.

> Would you be amenable to feature-gating these public helpers behind a `test-util` feature? This is a pretty common approach we've seen in other crates. That sounds reasonable to me....

Thanks for testing. Your observations match my understanding. Is the original issue valid?

Hi @fedetorres93, I am sorry for the delay. I appreciate the time you invested making a solid patch! As I assume many of us do, I am maintaining this crate...

Works for me. Thanks!

> encode time: [37.371 ms 37.433 ms 37.499 ms] > change: [+33.474% +34.031% +34.591%] (p = 0.00 < 0.05) > Performance has regressed. A 35% performance regression seems significant, especially...

> One thing to bear in mind is that, in the current version of the client, no validations are performed on metric or label names, so if I register a...

Corresponding draft section: > Clients can, however, send a SETTINGS frame, multiple WebTransport CONNECT requests, WebTransport data streams, and WebTransport datagrams all within a single flight. As those can arrive...

Note that the issue title includes both streams and datagrams that might arrive early. #2806 only fixes the issue for streams. I still think we should close here. Datagrams need...

@afrind yes, as long as the 200 OK is before the WebTransport datagram in the QUIC packet, the client will correctly handle both. https://github.com/mozilla/neqo/pull/2979 will add a test for it.