[ntp:questions] synchronization distance

Ryan Malayter malayter at gmail.com
Mon Dec 6 16:38:40 UTC 2010

On Sat, Dec 4, 2010 at 3:55 PM, Mike S <mikes at flatsurface.com> wrote:
> Your complaint is misplaced. NTPv4 is well defined, see RFC 5905 ( http://tools.ietf.org/html/rfc5905 ).
> The RFC states "In NTPv4, tau ranges from 4 (16 s) to 17 (36 h)." Just as with many previous things-MS,
> they've gone their own way, and are not following the specification. They simply don't play well with others.
> Not following the specifications (like using an improper addr-spec such as "david at ex.djwhome.demon.invalid"
> in an RFC 822 message) breaks things.

Look at the publication date of RFC 5905. There is no way MSFT
products that are older than that date could possibly be NTPv4
compliant. In fact, there is no way ANY implementation of NTP except
the reference implementation could possibly be NTPv4 compliant until
after that publication date.

MSFT claims only NTPv3 compliance for Windows Time Service, and only
for the version on Windows Server 2003 and later.

This "synchronize once a week" about Windows Time Service is
nonsense... in the last decade, I have *never* seen a Windows system
(even 2000 and XP) have a polling interval that long unless manually
configured to do so. The default configuration in Windows 2000 and XP
for *standalone* windows systems was to poll time.windows.com once
every eight hours. Systems that are part of a Windows Active Directory
domain poll local domain controllers more frequently,and have always
done so.


More information about the questions mailing list