Mat Trudel

Closing for hygiene

Those are some serious numbers! I'd love to see the approach taken for this - probably the best place to land initial work for this would be in a helper...

I'm *pretty sure* that we can guarantee (at least in Bandit's case) that an iolist is either a binary or a (possibly improper) recursively defined list of iolists. Extremely excited...

The [typedef](https://hexdocs.pm/elixir/1.12/typespecs.html#built-in-types) for iolists includes bytes: ``` iodata(): iolist() | binary() iolist(): maybe_improper_list(byte() | binary() | iolist(), binary() | []) byte: 0..255 ``` I suppose that means that all charlists...

> @mtrudel I noticed @fhunleth doing some interesting iodata experiments that might be relevant: > > https://github.com/fhunleth/iodata_nif Interesting! Frank's as sharp as they come. I'm certainly going to look at...

Any chance to have this integrated & released?

The error case[^1] telemetry events won't have many / most of the measurements, this is ~more or less by design since it simplifies the mainline code so much. That being...

Sorry for the late response! I think this may rhyme with #202 and #305. I've pushed up a branch that disables exit trapping on HTTP/1 connections (so they'll close as...

I've tightened up the semantics on this and pushed up a proper PR at #346. I'd *love* feedback to see if this fixes your issues @mickel8, @MrYawe, @hubertlepicki

There has been some movement with third party benchmarks in the past: https://github.com/mtrudel/bandit/issues/69. Unsure how timely that remains. We also have an open PR to improve the visibility of our...