for most HTTP calls, the request side will need a smaller buffer than the response side. Would it be interesting to allow switching to larger buffers, or using vectored IO in cases where we might transfer more data?
If we know the content length, or know that the content will be chunked, we could switch to another IO mode.
I'm not convinced yet we should switch to something like this, but it should be interesting to test
for most HTTP calls, the request side will need a smaller buffer than the response side. Would it be interesting to allow switching to larger buffers, or using vectored IO in cases where we might transfer more data?
If we know the content length, or know that the content will be chunked, we could switch to another IO mode.
I'm not convinced yet we should switch to something like this, but it should be interesting to test