Greg Smith

I still think you are focusing on the wrong behavior. I think(hope) the 99.44% use case is this: /* enable keepalives, use system defaults as they are, hopefully, appropriate for...

All I'm arguing for is the ability to use the system settings when enabling keepalives[1]. If you have delay < 1 return an error[2] then you've removed an important capability,...

I'll test it, if it works, I have no problem with that solution.

SIO_KEEPALIVE_VALS is worse. `uv_tcp_keepalive(handle, 1, 0)` always sends keepalives as fast as possible ( even on window 10/2019)

I think my original suggestion gets us to "behavior needs to be ... consistent across platforms". The change brings a few rogue OSes (2012r2/2016) inline with other platforms (tested: ubuntu...

I added doc to the PR, to hopefully clear up expected behavior