[ntp:questions] Re: Drift handling....
David J Taylor
david-taylor at blueyonder.co.not-this-bit.nor-this-part.uk.invalid
Mon Jan 2 09:40:28 UTC 2006
Harlan Stenn wrote:
> The issue is not "how often is the drift file written".
> ntpd calculates the drift on an ongoing basis, and uses that value
> Once an hour the value is written to the drift file.
OK, it's my phrasing.
I'm suggesting that ntpd not write a fresh value until the new one is
proven for at least 24 hours, and that perhaps ntpd allow for a perhaps
sinusoidal variation (or four six hourly samples throughout the day) if
this is required to deal with diurnal variations. My observations were
purely of the value written to the drift file.
Perhaps ntpd should be calculating on a 24-hour, not a 1-hour basis?
More information about the questions