[ntp:questions] Re: Contradicting RFC<---> Implementation

David L. Mills mills at udel.edu
Mon Oct 3 14:51:35 UTC 2005


First, the NTPv3 specificatino rfc1305 does not accurately describe 
NTPv4 reality; there are subtle differences. Best way to address the 
differences is the set of briefings and flow charts at the NTP project 
page www.eecis.udel.edu/~mills/ntp.html. MAXDISTANCE refers to the 
distance threshold above which the selection algorithm will not believe 
the particular source. The distance is equal to one-half the root delay 
plus the dispersion, which itself has several components. See the error 
budget in the architecture briefing for details.

The NTPv4 sanity checks are detailed in the flow charts. There are quite 
a number of them in addition to the checks detailed in rfc1305. See the 
fit() algorithm for a definitive summary of what is and is not 
acceptable for a given server.


Remo wrote:
> Hi,
> The NTP RFC does'nt mention anything explicitly about the sanity checks
> of the clock_select procedure, but the rootdispersion check for an
> association is made with the MAXDISPERSE(16 sec).{according to RFC}.
> In some implementations, I could see that the limit is MAXDISTANCE + 2
> (which is 3 seconds). In the time, wondering about this change and the
> "3 seconds", I found that the w32time service provides a rootdispersion
> above 1 second.
> Could anyone let me know the reason for this change and why the w32time
> service is way above threshold of 1 sec???
> Thanks for your time!!!
> Thanks,
> Remo

More information about the questions mailing list