[ntp:questions] Two servers, one client : one problem

Richard B. Gilbert rgilbert88 at comcast.net
Wed May 4 14:00:58 UTC 2011

On 5/4/2011 2:49 AM, David Woolley wrote:
> Richard B. Gilbert wrote:
>> NTP is hierarchical! It expects to get time from an external source
>> such as NTP servers on the internet. The hierarchy starts with an
>> "atomic clock" such as those operated by NIST or equivalent. These are
> There is nothing that prevents an atomic clock (no quotes) being used at
> any level in the hierarchy. Good means much better than the estimated
> error contributions between the standard and the ntp client.

There is nothing to prevent you from setting the stratum of your atomic
clock to; e.g. 12.

I fail to see any point in doing so.

> The key requirement, that is being violated here, is that the top
> stratum sources should be good approximations to the same time scale.
> "atomic clock" in quotes is marketing hype for devices that may coast
> with un-temperature controlled crystal accuracy for 24 hours, but then
> be resynchronised, within about 100ms to a radio transmission derived
> from an atomic clock calibrated to a national standard, which is in turn
> calibrated (after the fact) to UTC.

Calling something "an atomic clock" does not make it one!!!!!  I was 
referring, or trying to, *the real thing*.

Thank you for trimming the relevant parts of the past posts!

More information about the questions mailing list