[ntp:questions] Re: ntp client over satellite and no CMOS battery

Danny Mayer mayer at gis.net
Fri Oct 7 04:38:25 UTC 2005

David L. Mills wrote:
> Tim,
> A good deal of quality design time went into the statistical analysis of 
> errors and the manner in which statistics are presented to the 
> application, in this case via ntp_gettime(). You don't want to know 
> anything other than the carefully maintained maximum error and estimated 
> error statistics. Neither the stratum of the server nor the poll 
> interval nor the precision nor any other statistic is necessary to 
> evaluate the quality of time. The maximum error and estimated error 
> already take those factors into account. You really don't want to second 
> guess what the server is doing, just that it and the local client are 
> providing dependable statistics.
> As I said in my last, your clock is Big Ben and NTP is manipulating the 
> number of farthings, h'apennies, pence, shillings, pounds and guineas on 
> the pendulum. [Gawd, after 33 years since decimalization, I'm not sure I 
> remember how to spell all that coinage.]
> Dave

You did forget the threepence, half-crowns, crowns, and florins, all of 
which I have in my collection.

I seems to remember reading somewhere that Big Ben actually uses NTP to 
stay accurate.


More information about the questions mailing list