Subject: Re: todr changes to improve clock accuracy across sleeps &
To: Garrett D'Amore <>
From: Richard Earnshaw <>
List: tech-kern
Date: 09/08/2006 19:34:01
On Fri, 08 Sep 2006 07:24:14 PDT, "Garrett D'Amore" wrote:
> Perry E. Metzger wrote:

> > You miss #3
> >
> > 3) As with my laptop, sometimes the box is connected to the network
> >    and has access to external references, and sometimes it is not, and
> >    there is no easy way to predict in advance what state you will be in.
> >   
> In case 3, ntp will correct the drift when you connect.  While
> disconnected, if you did say 10 shutdowns/restarts a day, the worst case
> drift is about 20 seconds.  I know of onboard RTCs that drift that much
> per day anyway.   (Granted, that's pretty darn bad.)

No.  That's the whole point.  If you don't write the clock on each 
shutdown then the drift is zero.  At startup your machine will be anywhere 
within a second of real time (assuming your RTC doesn't drift); but since 
you never write that slightly erroneous value back to the RTC there's no 
systematic change and thus no compounding effect.