[ntp:questions] NTP vs chrony comparison (Was: oscillations in ntp clock synchronization)

Brian Utterback brian.utterback at sun.com
Mon Jan 28 13:14:05 UTC 2008

Danny Mayer wrote:
> I'm not sure what it was you were trying to modify, the NTP v4 draft, 
> the book, something else.
> I never said that you can consider ntpd in isolation. It's part of a 
> network of ntpd servers and each is interacting with others. That's a 
> very important part of the specification and each is affected by the 
> others.
> Danny

Not implying that you did. I think we are in violent agreement here,
Danny. My point is simply that the spec concentrates on the algorithms
needed to groom the system clock, and doesn't really deal with what
the requirements are to play well in an NTP network. This is
understandable, since those requirements are much less well defined.

My point is that intuitively, we don't expect wire specs to say much
about internal implementation details. However, with NTP we tend to
distrust anything that doesn't implement the algorithms used in the
reference implementation. Furthermore, the requirements for being a
client are obviously different from being a server, as reflected
in the SNTP spec. But the spec doesn't really deal with the
characteristics of a good server in a way that you could implement,
it just gives specs on a good client and says implement that if you
only want a client, and implement the whole package if you want
a server.


"Mr. Jefferson, build up that wall!"
Brian Utterback - Solaris RPE, Sun Microsystems, Inc.
Ph:877-259-7345, Em:brian.utterback-at-ess-you-enn-dot-kom

More information about the questions mailing list