Accurate time without a network connection?

[Joel Rees]

> On Tue, Apr 22, 2003 at 03:26:19PM -0400, Charles Swiger wrote:
> > 
> > Ok.  If you run NTPD with only the local hardware clock for a 
> > reference, wait for a week, and then see how the intrinsic drift of the 
> > hardware compares with "real time" (using your watch or some other time 
> > source), you can adjust /etc/ntp.drift by hand.  This isn't going to be 
> > perfect, but it's going to be much more accurate than doing nothing.
> 
> Good. But already tried that. The situation is multiple systems have to
> run with something near the same time, but no bidirectional contact. And
> need to operate for years. Letting ntpd tune itself and then free run
> works much better than the system clock alone but only good for weeks,
> not months.
> 
> As for exactly what time the systems have, it doesn't much matter as
> long as they all have the same time.

If you're serious about it not mattering exactly what time they have,
what's the problem with letting one machine be the time server, letting
it tune itself and then free run, and syncing all the rest to the
slightly-off-time-server?

-- 
Joel Rees <joel at alpsgiken.gr.jp>