4 ms·
This is not really a problem. Some thoughts: 1. Time is never the same on two different computers. 2. Time is different when you send it and receive it (speed
by batou 11y ago
This is not really a problem. Some thoughts:
1. Time is never the same on two different computers.
2. Time is different when you send it and receive it (speed of light).
3. Time isn't necessarily linear (time dilation).
4. Clocks drift, even with NTP.
5. Some systems are isolated enough to accept inaccurate human input and time synchronisation.
Time is an invention for human consumption mainly. Machines are better served with explicit synchronisation (paxos, distributed transactions) or not at all (eventual consistency). Examples in brackets.
Time, if we lose a few seconds here and there, meh, don't sweat it.
- xyzzyz 11y agoMachines are better served with explicit synchronisation (paxos, distributed transactions) or not at all (eventual consistency) Using atomic clocks to sync distributed systems is actually done in practice. http://static.googleusercontent.com/media/research.google.com/en//archive/spanner-osdi2012.pdf http://static.googleusercontent.com/media/research.google.co...
- batou 11y agoUtterly insane IMHO. If there's a single error in any temporal data, what happens? Even relatively reliable frequency/time standards (HP/Agilent come to mind) aren't 100% available.