[time-nuts] Divide by five ->Ensemble
kb8tq at n1k.org
Mon Nov 10 07:14:55 EST 2014
> On Nov 10, 2014, at 1:59 AM, Poul-Henning Kamp <phk at phk.freebsd.dk> wrote:
> In message <4CDC0090-DD4B-4380-ACC1-40A80D3BC20C at n1k.org>, Bob Camp writes:
>> Here’s what I’m saying:
>> The NTP algorithm as it is written and as it is implemented results
>> in an output clock that does *not* improve when a number of very
>> good clocks are being used [...]
> The crucial word here is "good clocks".
> A "good clock" in NTP is nearby and because it is nearby, the chances
> of several of them overlapping is usually not high enough for any
> clock combining to happen.
Yes, in normal NTP land it’s not in any way worth worrying about.
> The NTP algorithms were built for 10-100msec packet delays, not
> low microsecond packet delays.
NTP is designed to do something very different.
> Poul-Henning Kamp | UNIX since Zilog Zeus 3.20
> phk at FreeBSD.ORG | TCP/IP since RFC 956
> FreeBSD committer | BSD since 4.3-tahoe
> Never attribute to malice what can adequately be explained by incompetence.
More information about the time-nuts