In 1998 when I started my career in Novell in 1998 I started work in
time synchronization.
I wrote a protocol convertor between Novell Timesync and UNIX NTP and
this work saved
Novell a 5m dollar British telecom deal and I also did another bugfix
for which I got an award.
Of course in big MNCs awards are one way to manage political equations. ;)
Now time synchronization is not something that is easy. Not at all.
A prof in univ of Delaware Dave Mills wrote the NTP rfc and he
maintains the various other
literature on it.
The documentation is incredibly chatty but very comprehensive. You
obtain atomic clock precision
from Cesium crystals aboard satellites.
There are various NTP stratum 1 servers around the world that fetch
this time through radio long wave (sat is microwave)
and also transmit in radio.
Of course with computers we only have to use UDP. ;)
There is very simple basic school algebra here. You get time from a
remote stratum 1 or 2 NTP server and your time is
off by a few seconds, and then you have to account for the latency of
the packet transit.
It can be corrected provided the uplink and downlink delays are identical.
Of course all this was in 1998.
Subsequently a German Henning Brauer wrote OpenNTPD(www.openntpd.org)
for OpenBSD.
That removed all the nonsense and complexity of the UNIX ntpd or xntpd toolset.
Now this is installed in all OpenBSD machines as this is in base.
In fact you can get accurate time if you use my
http://liveusb-openbsd.sf.net
Try it. It is incredibly useful to get started.
Or try this livecd.
http://livecd-openbsd.sf.net
A simple way to correct the skew or some timezone issue just this:
# rdate ptbtime1.ptb.de
I do this when my computer is off by a large amount.
The RTC clocks that run with the CMOS clock on the mobo are usually
not accurate.
Not even as good as the wrist watches you wear.
Why is time imp?
Well all processes depend on wallclock time.
-Girish
--
Gayatri Hitech
http://gayatri-hitech.com
