[ntp:questions] NTPd instability and openntpd

Harlan Stenn stenn at ntp.org
Wed Sep 19 01:19:21 UTC 2007


>>> In article <nemoMon091707113110 at news.fastwebnet.it>, Mij <mij at bitchx.it> writes:

Somebody wrote:
>> There are issues with OpenNTPd and I believe it does not support mode 6
>> packets which is required to run ntpq against it. In addition it is notan
>> NTP server but an SNTP server.

mij> It's a SNTPv4 + NTPv3 server.  From
mij> <http://www.openbsd.org/cgi-bin/man.cgi?query=ntpd:>

mij> "ntpd can also act as an NTP server itself, redistributing the local
mij> time. It implements the Simple Network Time Protocol version 4, as
mij> described in RFC 2030, and the Network Time Protocol version 3, as
mij> described inRFC 1305."

Just because something *can* be done does not mean it *should* be done.

In particular, an SMTP implementation should (must?) only distribute time if
it is talking directly to a time source.

mij> Reading from
mij> <http://sunsite.ualberta.ca/Documentation/Misc/ntp-4.0.99a/release.htm>
mij> the major additions of NTPv4 wrt NTPv3 involve only advanced features
mij> (autokey authentication, 64bit, multi/many cast) that are pretty
mij> uselessin the ntp pool.

They didn't talk about bugfixes or significant algorithmic improvements.

mij> I'd say the major concerns in pool are accuracy and
mij> compatibility/availability. I see that NTPv4 clients understand NTPv3
mij> reponses, so openntpd would be in service for new clients. Moreover,
mij> from real data this is how myserver responds

There should be no compatibility issues with any version of an NTP client
talking to any NTP server.  The protocol handles differing versions.

SNTP clients are designed for "leaf" nodes.  They SHOULD (and it may be a
MUST) not be used to distribute time unless they are talking directly to a
reference clock.

H




More information about the questions mailing list