Xref: utzoo comp.unix.sysv386:9038 comp.unix.wizards:25992
Path: utzoo!utgpu!news-server.csri.toronto.edu!bonnie.concordia.ca!thunder.mcrcim.mcgill.edu!snorkelwacker.mit.edu!mintaka!spdcc!rbraun
From: rbraun@spdcc.COM (Rich Braun)
Newsgroups: comp.unix.sysv386,comp.unix.wizards
Subject: Re: Network Time-keeping
Message-ID: <7904@spdcc.SPDCC.COM>
Date: 17 Jun 91 16:27:12 GMT
References: <1991Jun14.024635.15860@cbnews.cb.att.com> <14979@ulysses.att.com>
Followup-To: comp.unix.sysv386
Organization: Kronos Inc., Waltham, Mass.
Lines: 19

For a couple of months, I had a network time server running which actually
kept accurate time (in file /etc/tcp, uncomment the line which starts up
'timed -M' on *one* of the systems, and just run 'timed' on others), but
as soon as a second machine was brought up running 3.2.2, the clock started
losing time (similar to the way it does if two machines are brought up
with 'timed -M').

I'm about to give up on synchronization myself and just write a script
for each system's startup file which just does 'set rsh <server> date'
followed by 'date $1 $2 $3' or some such kludge.

There seems to be no way to win by merely using software shipped by SCO.
Software can be downloaded from the University of Delaware and other
places, but I'm not yet convinced that's necessary.  My network presents
a bit of a problem, because it contains onesies and twosies of systems
from various vendors, making it rather difficult to compile any software
which has to run on all systems.

-rich