Any data on HTTP/3 performance? I don't see it in the book. There's the general claim that it's faster/lower latency, but there are no numbers behind that claim -- last time I checked QUIC's performance benefits were incredibly slight.
It is really easy to observe the performance benefit of QUIC in a congested datacenter network. In the face of loss, QUIC tail latency is dramatically better than TCP tail latency. This is mainly due to TCP's 200ms minimum retransmit time; a single dropped packet will add at least 200ms to the request time (modulo tail loss probing which can lower this to ~20ms in many cases). When your request service time is 10µs this makes a huge difference.
