Doubts about discarding monotonic time

[jech]

Doubts about discarding monotonic time

[paulwe]

in practice i don't think it matters. on any well behaved system the aberrations in wall clock time should be smaller than the tolerance for timeout/keepalive scheduling.

because it's part of the public api we can't really do anything more clever. go ahead and revert it if you like. none of this had a measurable impact on my target metric.

[jech]

in practice i don't think it matters. on any well behaved system the aberrations in wall clock time should be smaller than the tolerance for timeout/keepalive scheduling.

What happens if the system clock is stepped?

Some systems have no battery-backed clock. On such systems, the system clock is stepped after NTP has had time to converge, typically a couple of minutes after boot. If I understand your patch correctly, it will cause Pion to break after the clock is stepped.

none of this had a measurable impact on my target metric.ou

Did you test while stepping the clock?

go ahead and revert it if you like

@Sean-Der could you please consider PRs #698 and #699? They unbreak Pion on systems with no battery-backed clock (such as Raspberry Pis and other SBCs).

[paulwe]

sorry i thought you had commit or i would have done it myself. like i said, this didn't fix the issue i was trying to solve so i'm not super attached to it.

[jech]

Thanks!