"David Woolley" <[email protected]> wrote in message
news:[email protected]...
[]
I think you miss the point.
Quite possibly, it des get a bit esoteric at times!
ntpd does better because the disciplined clock time is about an order of
magnitude more accurate than the spread of offsets, whereas the ntpdate
time jumps around with the same order of magnitude as the offsets.
There is no way of measuring the true ntpd time keeping error using just
ntpd, because any part of it that could be measured should be being
removed by the time discipline daemon.
Your graphs are not graphs of the ntpd disciplined clock errors, but
rather they are graphs of the sample errors.
So, if I understand you correctly, you are saying that because ntp
measures and reports an offset, it will try and correct the clock in the
opposite direction, so that the actual clock error is likely to be less?
I can't get at any better value, so the plots are currently the best I can
do. I label them as "offset" since that's the field they come from.
I have no interest in using ntpdate, as I understand it to be a less
accurate approach.
Thanks,
David
_______________________________________________
questions mailing list
[email protected]
http://lists.ntp.org/listinfo/questions
