[ntp:questions] nmea and initial large offset

David Lord snews at lordynet.org
Sun May 9 17:34:34 UTC 2010


Rob wrote:
> unruh <unruh at wormhole.physics.ubc.ca> wrote:
>>> Maybe the chrony people should have started with a fork from ntpd.
>>> Then I would probably be using it now.
>> I doubt it. chrony was largely written by one person, Richard Curnow.
>> His philosophy on how to control clocks was completely different from
>> Mill's and has been shown to work much better. It has now been taken
>> over by a small group with M Lichvar contributing most to the
>> development. It is a well developed program, different from ntpd in many
>> many ways, and experimentally much better in many many ways. It needs
>> theoretical work (eg are there conditions in which it becomes unstable?
>> Are there conditions under which it performes worse than a simply
>> Markovian feedback device that ntpd uses? Are there conditions in which
>> the clock filter algorithm of ntpd might be useful, and how can one
>> recognize them automatically. etc) and implimentational ideas. 
>>
>> But as to you using it, I have no idea why, if it were called a fork of
>> ntpd you would be more likely to use it. 
> 
> I have heard that it does not support local reference clocks.
> I have a DCF-77 receiver and a GPS receiver which is interfaced via gpsd.
> My understanding is that chrony cannot use those, and uses only other
> servers via the network.
> That is not very useful for me.
> When it had started as a fork from ntpd, this issue would probably not
> exist (because ntpd supports both my local reference clocks).

Recent version of chrony has some refclock support. I've only
tried 1.24-pre1 on NetBSD and there were too many problems
for me to be able to sort out any fix. It should be ok on both
FreeBSD and Linux. Support for reference clocks (SHM, SOCK, PPS
drivers). So both DCF-77 and GPS are covered via SHM driver +
radioclkd and gpsd.


David




More information about the questions mailing list