Benjamin Saunders

Haven't looked at the PR yet, but initial thoughts from this overview: Transport parameter and frame IDs should be reserved through the IETF before merging; use whatever you like for...

This looks good to me overall. > Initial attempt for linux. Not sure what to do about Windows. I think Windows should look very similar to Unix here, with easy...

You're pulling in `nalgebra_lapack`, but you don't seem to actually be using it for anything.

`euler_angles_ordered` should probably internally clamp into the valid range. In general, any rotation matrix (or other normalized structure) you might handle is going to be prone to that type of...

> I really don't expect to encounter NANs (or having to check the validity of each -- possibly hidden -- intermediate result manually) if I don't use _unchecked in my...

Doc PRs would be very welcome! This is the behavior throughout nalgebra. [`Rotation3::renormalize`](https://docs.rs/nalgebra/latest/nalgebra/geometry/type.Rotation3.html#method.renormalize) and [`Unit::renormalize`](https://docs.rs/nalgebra/latest/nalgebra/base/struct.Unit.html#method.renormalize) may be useful references.

As I said in https://github.com/dimforge/nalgebra/issues/1481#issuecomment-2646655852, that's the responsibility of the logic that has that constraint, i.e. `euler_angles_ordered`.

It looks like you're comparing connection-level stats in iperf3/secnetperf to stream-level stats in Quinn. You also don't seem to be using the default `perf` configuration. What is the exact command...

Thanks for the update! I'm surprised that getting rid of the pipeline bubbles had so little effect. Maybe the original per-stream statistics were already excluding that well enough. Another area...

Well, that confirms it was blocked at two specific points in time. We really need to chart/aggregate state over time to draw meaningful conclusions -- for example, compare time spent...