[ntp:hackers] A wee experiment

David L. Mills mills at udel.edu
Tue Dec 2 08:22:12 PST 2003


Reg,

Most of the statistics fields in ntpdc are calculated per NTPv3 and are
in error with NTPv4. The dispersion field is available in ntpq, but not
in the condensed pe billboard. If what you want is the expected error,
then the jitter estimate is far better than dispersion. However,
whatever winds your clock is good.

Dave

clemens at dwf.com wrote:
> 
> > Reg,
> >
> >
> > Aside, lose the ntpdc. It is broken, evil and just plain wrong in many
> > places. Trust ntpq verily.
> >
> > Dave
> >
> 
> Yes, Yes, I know, but i like the wider fields in ntpdq, and I like the
> disp field rather than the jitter field.  The disp field seems to walk
> down nicely towards zero when things are working, and there seems to
> be no coorespondence between the disp and jitter when I print out both
> ntpq and ntpdc.
> 
> With wider fields (I think this is just a formatting thing) and disp
> Id shift to using ntpq in an instance.
> 
> --
>                                         Reg.Clemens
>                                         reg at dwf.com



More information about the hackers mailing list