[ntp:questions] Re: Can sychronize with a slightyly different NTP versios

David Woolley david at djwhome.demon.co.uk
Sat Mar 4 11:51:06 UTC 2006


In article <Ud2dnSDB26ryQJXZnZ2dnUVZ_vidnZ2d at comcast.com>,
Richard B. Gilbert <rgilbert88 at comcast.net> wrote:

> With an unsynchronized local clock as a server, how much could he care 
> about phase and/or frequency error?  Ntpd can hold a client in really 

People who don't care about these parameters generally don't ask how
to synchronize their clocks.  There seem to be a lot of people who have
a requirement that multiple machines track each other in clock phase
(and in frequency to the extent that frequency is the derivative of
phase) to a high degree of accuracy, but don't have a requirement to
track a global standard accurately.  There are even protocols, in parts
of the Unix world, that specifically cater to this, such as timed.

Such people may well find that the sharp phase jumps and only 1 second
resolution associated with rdate are intolerable, and, at reasonable 
polling rates, the size of each correction can also be too large.

Typically they also do not have access to budgets for external services
or the purchase of hardware, even though the opportunity cost of 
implementing by other means may be larger than the hardware costs.
tight synchronization (100 usec or better) IF (big IF) it has a stable 
reference; e.g. a server synchronized to a GPS reference can synchronize 
clients within a few microseconds.  I've never had reason to try to 
synchronize to an unsynchronized local clock but I would not expect 
frequency or phase error levels to brag about.




More information about the questions mailing list