undefined | Better HN

0 pointsyjftsjthsd-h5y ago0 comments

I was under the impression that the OS did only track time in seconds since the epoch, and whenever you see something else it's just the software translating for human convenience. If that's the case, then it this doesn't really make sense; humans will want something friendlier than epoch time on Mars, too.

0 comments

2 comments · 1 top-level

bonzini5y ago· 1 in thread

Sort of. Unix time does not count leap seconds, all days are 86400 seconds long in Unix time. So the actual number of seconds since the epoch is a bit larger than what time() returns.

I don't know how Windows deals with leap seconds (its epoch is in 1601 or something like that). Edit: it doesn't, the clock goes 1 second ahead of UTC and it's then fixed by NTP.

pedrocr5y ago

So time() is actually days since epoch with 1/86400 precision. That makes the Google solution to smear leap seconds with NTP instead of making them explicit make even more sense.

j / k navigate · click thread line to collapse