Hacker News Reader

Comment by sfblah | Hacker News Reader

It mentions the days being longer, but what about the years? What even constitutes a year on Mars? Seems like more thought needs to go into how this would work if/when we become interplanetary. Assuming knowing where you are in the year has any value on Mars, we'd want some way to track that. I wonder also how this would work on the moon.

View on HN

My background is physics and during an extended beer session my group talked extensively about what properties an interplanetary time keeping system would need to have.

The answer we came up with was: thank god we'd be retired or dead by the time it became a problem that we had to deal with day to day.

You would need something like international atomic time to be the basis for it, but unlike TAI which is the weighted average of some 400 atomic clocks on the surface of the planet, you would need to define it with respect to free space (which we can't access) and with respect to a single event that's accessible to everyone yet human scaled enough that it happened in written history (the best we came up with was the first nuclear blast).

Then you would need to have a calculation that turned it into local time which would be completely decoupled from any actual time keeping. Think printing the result of a square root in Roman numerals vs trying to do the calculation in Roman numerals.

Unix epoch is laughably inadequate for anything like that. Basically we have nothing today and you'd need an international agreement to have a sane time keeping system going forwards.

One anticipates a network that operates with a faster-than-light "because" pipe, or you just have a planetary local NTP server and let something upstream do the Einsteinian gymnastics with the timestamps.

What's wrong w/ TCB (https://en.wikipedia.org/wiki/Barycentric_Coordinate_Time) and TDB (https://en.wikipedia.org/wiki/Barycentric_Dynamical_Time)?

Judging by DST and the gregorian calander I'm sure we'll end up with a system that doesn't even approach the utility of what you're thinking.

Everyone will probably just use Eastern Time on mars complete with Daylight saving, maybe TAI if we're lucky.

Probably best to just drop the notion of "days", "years", etc. entirely, and instead just track the number of kiloseconds, megaseconds, etc. since the Unix Epoch.

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.

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.

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.

Since the Unix Epoch where? Isn't being on another planet going to have to deal with the relativistic passage of time?

In the novel "A Deepness in the Sky" by Vernor Vinge, the characters tell time this way. I had to do some math early in the book to get a feel for what 1Msec & 1Gsec was.

In addition, they also keep track of their cold sleep time, because if you want to meet your spouse 50 years from now, you want to try and keep the same relative age.

France used the decimal time for a few years beginning in 1792 during the French Revolution. It didn't go well.

Reply on news.ycombinator.com