codspeed benchmarks show performance regressions if you click through to the report.
Alternatives
Maybe we want an Option<oneshot::Sender> field in RequestWrapper instead?
One important case to consider is the tee transform.
It needs to dynamically handle swapping between two different chains as the source of truth.
So when one chain is activated it should be the one to be capable of closing the client connection.
I think mock out the API of this approach and then compare against this PR.
prereq for https://github.com/shotover/shotover-proxy/pull/1715
codspeed benchmarks show performance regressions if you click through to the report.
Alternatives
Maybe we want an
Option<oneshot::Sender>
field in RequestWrapper instead?One important case to consider is the tee transform. It needs to dynamically handle swapping between two different chains as the source of truth. So when one chain is activated it should be the one to be capable of closing the client connection.
I think mock out the API of this approach and then compare against this PR.
If https://github.com/shotover/shotover-proxy/pull/1719 yields good performance I think we can create a
And then include it in RequestWrapper as a
&RequestMeta