[ntp:questions] unix and windows environment sync to one NT4 timeserver "HOW"

David L. Mills mills at udel.edu
Mon Aug 4 02:55:00 UTC 2003


Sometime earlier in these messages I claimed Win/XP did in fact work 
more or less correctly with a NTP client. Not quite. Win/XP updates the 
clock about once per week. But, after a couple of days the dispersion 
increases to something over the one-second upper limit and the NTP 
client will no longer believe it unless tinkered otherwise.

Having said that, I conclude Win/XP has lifted the NTPv3 protocol in 
that the leap bits are set to 11 (unsynchronized), apparently after two 
days. The NTPv4 server will not do that, since once the leap bits are 
set to something over than 11, they will never be set to that value 
again. To inidcate the server has lost synchronization, the stratum is 
set to zero. But, this doesn't happen until after eight polls have not 
been answered. Interestingly enough, if Win/XP followed the NTPv4 rules, 
a poll interval as large as a week would in fact work, enven though a 
much more satisfying interval would be one day.


David Woolley wrote:
> In article <bga0sl$qqn$1 at dewey.udel.edu>, David L. Mills <mills at udel.edu> wrote:
>>Now the bad; maybe not even bad at that. The same Win/XP Professional 
>>Edition machine does NOT return packets from a configured ntpd client. 
> My impression was that the original versions of Microsoft's SNTP
> approximation returned certain fields from the request that would
> normally be used to determine the health of the server.  Thus, an already
> synchronized client would see its own good status and consider the
> server good, but an unsynchronised one would see it as bad.  I haven't
> rechecked this.

More information about the questions mailing list