[time-nuts] Does GPS time reception work everywhere all of the time?

[Poul-Henning Kamp]

In message <201008310942.o7V9geRF025941 at stenn.ntp.org>, Harlan Stenn writes:
>You wrote:

>> The problem is that if multiple servers with roughly identical
>> performance survive the culling, the clock selection codes
>> commitment-anxiety makes i switch partner far too often.
>> 
>> If you allow the automatic poll-rate tweaker into the game, instead
>> of sensibly clamping maxpoll, it gets even more interesting.
>
>Has this been brought to DLM's attention?

Yes, ages ago.  I belive the result was a documentation note that
too many partners were a bad idea.

>And I assume this is not a problem with your code because you code is
>designed for S1 use as opposed to higher-stratum systems that are in a
>better "position" to have multiple servers, right?

Remote servers in NTPns usually serve only as sanity-checks or fallbacks
and the selection is purely administrative (#votes)

-- 
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.