Unruh wrote: > > No, it was last synchronized from IRIG then. Since then it is supposed to > have been synchronized from the H masers they keep as a primary world time > source for Neatherland's UTC time source. I assume this is being done by
If they are only synchronising to IRIG every 3+ days, the root dispersion is correct, and 52ms is really quite a good actual accumulaed error (root dispersion is a pessimistically high value, assuming that the clock was originally properly synchronised, and there haven't been any anomalies of the sort associated with clock calibration on recent versions of Linux). > something other than ntp (eg a separate oven controlled clock source driven > by the world TAI standard)and thus ntp does not see it, and thinks that the > last time the local clock was synchronized was a week ago, while actually > the local clock is continually synchronized to far far better than ntp > could ever hope to do it. But noone told ntp. If the local clock is being synchronised by means other than ntpd, you have the original, and most valid reason for using the local clock driver. The local clock driver's fault with root dispersion is that it actually gives a wildly optimistic value, by resetting it every poll interval, which then fails to reflect the actual error on an undisciplined local clock (you seem to be speculating on a disciplined, just not by ntpd, one). Basically, if the local clock is being synchronised directly, they should be running the local clock driver, as preferred, and with a low stratum (0 if they have some means of promptly shutting down if synchronisation fails). Generally, we are guessing at their real configuration, but either their configuration or ntpd is broken, as far as acting as NTP server is concerned. However, it is likely to still be acceptable to w32time, and if that is its intended purpose, it may well be OK, even if one would question a starndards organisation providing such a low quality service, even if the clients only need low quality. _______________________________________________ questions mailing list [email protected] https://lists.ntp.org/mailman/listinfo/questions
