You have piqued my curiosity. I don't experience a great amount of time
drift as my computer is constantly being cycled back and forth between
Linux and Windows. I have a utility that runs every time I start
Windows that fetches the correct time from the atomic clock in Boulder,
Co.
I know a certain amount of time drift is inevitable from the theory I
have studied. How much is excessive I don't know for sure. If it helps
the issue I have posed the question to other sources. I will let you
know what the concensus is when I get some answers back.
Ken Wilson
First Law of Optimization: The speed of a nonworking program is
irrelevant
(Steve Heller, 'Efficient C/C++ Programming')
> -----Original Message-----
> From: [EMAIL PROTECTED]
> [mailto:[EMAIL PROTECTED]]On Behalf Of Max Klohn
> Sent: Thursday, September 16, 1999 12:16 AM
> To: [EMAIL PROTECTED]
> Subject: RE: [expert] 6.1 Helios observations -time drifts-
>
>
> Le jeu, 16 sep 1999, Ken Wilson a �crit :
> >Time drift is not necessarily a bug but a normal fact of life in
> >computing. It may be the product of numerous interrupts. The more
> >processes you have running over a period of time the more
> opportunities
> >for interrupts to occur, the greater the time drift. That
> is why timed
> >exists, and also utilities that will periodically fetch time from a
> >reliable source, i.e. the atomic clock in Boulder, and set
> the current
> >time for the box holding the timed server and serving the rest of the
> >network.
>
> Well, something definitely "changed" since I upgraded to
> Mandrake. I never had
> such wide time drifts with RH 5.2. With 4 weeks uptime, my
> machine is off
> by one hour.
> I dont't know how to further trace the problem. This is why I
> am asking the
> list.
> >
> >We are getting into way too much bashing. This thread
> should really be
> >named '6.1 Helios observations (Bug reports the sequel)'.
> In deference
> >to Gael I think we should be 100%, or as close as possible, sure
> >something is a bug and not just our own error or failure to get up to
> >speed on a process/problem before reporting same.
>
> I am 100% sure of this one. SAME hardware before + the very
> same daemons
> running with RH 5.2=no time drifts. I have now three servers
> up and running and
> they all drift. I do not whish to purchase a serial interface
> to a radio clock
> (which BTW could come more expensive than another distro)
> until I am resonably
> convinced this is not a software glitch.
>
> I am willing to stick with Mandrake, especially since getting
> used to a new
> distro's quirks is somewaht a PITA, as long as I see some
> issues are being
> addressed.
>
> Now how could I tackle this? I could run a cron job to
> resynch the clock with
> the hardware clock, but isn't the whole purpose of the
> software clock to be
> supposedly more precise than the hardware one? Any other
> (better) ideas?
>
> >
> >Maybe we could all drop this thread until we see what the
> actual release
> >turns out like.
>
> I definitely disagree with this. I think we need answers
> _before_ the new
> release if we want it to be better and more stable. Without
> this, releasing is
> just riding an ephemere fashion wave. The "distro of the day"
> kind of users will
> fade away, what remains, is the quality (or lack thereof) of
> the fixes &
> patches.
>
> MK
>
