Benjamin Saunders

This gets further now, but still fails before doing any testing: ``` > ./build.sh [...] [+] Building 33.1s (12/12) FINISHED docker:default => [server internal] load build definition from Dockerfile 0.0s...

Can you elaborate on the use case, and cite any prior art? We've already got an awful lot of variations on "write".

This patch to rustls fixes 3/4 failing tests: ```diff diff --git a/rustls/src/msgs/deframer.rs b/rustls/src/msgs/deframer.rs index 6da2d9d4..5d9d1fdd 100644 --- a/rustls/src/msgs/deframer.rs +++ b/rustls/src/msgs/deframer.rs @@ -253,7 +253,7 @@ impl MessageDeframer { // We're joining...

This fixes the remaining test: ```diff diff --git a/quinn-proto/src/tests/mod.rs b/quinn-proto/src/tests/mod.rs index f809ba1c..1b3b28de 100644 --- a/quinn-proto/src/tests/mod.rs +++ b/quinn-proto/src/tests/mod.rs @@ -379,9 +379,15 @@ fn reject_missing_client_cert() { let key = PrivatePkcs8KeyDer::from(CERTIFICATE.serialize_private_key_der()); let cert...

Made a deep pass; looks like only some minor cosmetic issues left.

On review, handling multiple connections on a single task might actually be undesirable as it prevents tokio's work-stealing from redistributing work when e.g. a single connection involves disproportionately large amounts...

I'm hesitant to expose these, because they aren't generally useful to well-behaved applications. CIDs are transient and routinely change mid-connection, and especially a remote CID (i.e. the one used for...

In a literal sense, yeah, but they also don't have a predictable relationship to the CIDs of post-handshake packets, which makes their usefulness limited.

> there is no corresponding way to retrieve that information in a server using quinn Normally, CIDs are opaque except to the party that generated them, so accessing CIDs from...

Closing for lack of activity, and because more graceful alternatives have been presented here and elsewhere. Feel free to open a new issue or PR to discuss if the best...