[ntp:questions] nmea and initial large offset

Richard B. Gilbert rgilbert88 at comcast.net
Sun May 9 12:13:45 UTC 2010


Rob wrote:
> Richard B. Gilbert <rgilbert88 at comcast.net> wrote:
>> Rob wrote:
>>> Richard B. Gilbert <rgilbert88 at comcast.net> wrote:
>>>> Please feel free to design your own algorithms to deal with the "systems 
>>>> not running 24h/day" and, while you're at it, deal with temperature 
>>>> variations and support drift rates up to 5,000 PPM, etc, etc.
>>> I don't feel free to design those because I can anticipate that the
>>> changes will be rejected and will not go into the distribution version.
>> Of course they will not go into the distribution version.  At least it 
>> won't go into Dave Mills' distribution.
>>
>> IF you can make it work across the full spectrum of NTPD usage it will 
>> be distributed somehow and people will use it if they perceive that it 
>> does, and does well, things that the official *Mills" distribution cannot!
> 
> I don't understand why certain features cannot be available as configurable
> options, so that mr Mills can have his limited functionality version while
> the rest of the world enjoys the practical, working, featureset.
> 
> For example, an option to tell ntpd that it should believe the reference
> clocks when they majority-vote that the correct time is very far from the
> OS time.  Like -g but also for local reference clocks like GPS, e.g.
> via ntpshm.
> 
> This should be possible.  And it should not affect mr Mills' systems
> when he makes sure he does not enable that option, so they happily
> continue to run in January 1970 after a RTC power loss when he likes that
> better.
> 
>> If Dave Mills says "It can't work!" I think the burden of proof that it 
>> can and does work is on you.
> 
> That is the problem.  He keeps referring back to his books and the
> facts he has collected 20 years ago, and a changing environment is not
> part of his world.

Dave's math is beyond me but I'm willing to take it on faith.  It works!

I suspect that faster convergence would carry a price such as large 
amplitude oscillation!

If you can't stand it, roll your own!  You may find that it's far more 
difficult than you think!




More information about the questions mailing list