Anton Firszov

Yeah I meant `c`.

@tmds Uploaded some results + traces to the Teams subfolder `RawSockets-0408`. `epoll` is done with `a=true`. The proof is missing from those results, but I hadn't noticed any difference with...

/azp run runtime-libraries-coreclr outerloop

/azp run runtime-libraries-coreclr outerloop

Approved over email. All test failures are unrelated.

> CPU spinning for a really long time as it reads in 8M scan lines. Ah yeah, when pushing for `AllocationLimitMegabytes` instead of dimension limits I didn't consider that reading...

IMO standardization should happen in parallel with the implementation. This is what we did with the HTTP metrics and we found many dragons. The runtime work is tracked by https://github.com/dotnet/runtime/issues/85372...

> Not sure what the `error.type` attribute would be in this case In SocketsHttpHandler/HttpClientHandler it is the namespace-qualified exception name in this case, ie. `System.Threading.Tasks.TaskCanceledException` or `System.OperationCanceledException`.

> Is this case handled in `http.client.request.duration` metric in .NET8.0? Yes, see https://learn.microsoft.com/en-us/dotnet/core/diagnostics/built-in-metrics-system-net#metric-httpclientrequestduration.

Note that `TaskCanceledException` is more common to be thrown by `HttpClient` in practice. If you think that will not increase confusion, I'm fine with `System.OperationCanceledException`, agree that it's more universal.