4.0.72 will be released Real Soon Now.

[Harlan Stenn]

As soon as Dave puts it on ftp.udel.edu, in either /pub/ntp/ or perhaps
even /pub/ntp/testing/, there will be:

ntp-4.0.71-4.0.72.diff.gz
ntp-4.0.72.tar.gz
ntp-4.0.72-export.tar.gz

Bugfixes, and probably other stuff that isn't documented (several
hackers don't update the ChangeLog).

OBTW, I'm using automake-1.2d (soon to become 1.2e, most likely).

I forgot to wrap the new required automake patches for this release - if
you want then let me know.

From the NEWS file:

(4.0.72)
* K&R C compiling should work again.
* IRIG patches.
* MX4200 driver patches.
* Jupiter driver added.
* Palisade driver added. Needs work (ANSI, ntoh/hton, sizeof double, ???)

[Kristofer T. Karas]

Harlan Stenn <st...@whimsy.udel.edu> wrote:
>As soon as Dave puts it on ftp.udel.edu, in either /pub/ntp/ or perhaps
>even /pub/ntp/testing/, there will be:

> ntp-4.0.72.tar.gz

>Bugfixes, and probably other stuff that isn't documented (several
>hackers don't update the ChangeLog).

Ah, but does it work with OSes *other* than Slowlaris?

I'm currently using 4.0.71a (which incorporates the loopfilter
workaround Dave Mills suggested for 4.0.71) in a test environment, and
sadly must report that, thought the FLL-disable problems are no longer
present, the PLL is barely able to keep time. Basically, here's what
I observe:

On every poll of the servers, xntpd-3.5.x makes a minute adjustment to
the frequency so as to keep the kernel clock spot-on; indeed, I rarely
see offsets of more than a few milliseconds (and this is on a heavily
loaded Linux machine too). In stark contract, ntpd-4.0.71a rarely
adjusts the frequency, perhaps only once every few hours, resulting in
offsets of anywhere from a few milliseconds to tens of milliseconds
(and on an almost unloaded server machine, for shame).

I have loopstats files available if anybody would like to see the
results of running 4.0.71a under Linux 2.0.33. I would love to be
able to run 4.0.7x on my production ntp server machine, but have to
back down to 3.5.92 until the extremely poor synchronizations are
solved.
--
Kristofer Karas * Senior systems engineer/SysAdmin
AMA/CCS DoD RF900RR HawkGT !car * BI Deaconess Medical Center, Boston
"Build a system that even a fool can use, * http://ktk.bidmc.harvard.edu/~ktk/
and only a fool will want to use it." * Will design LISP machines for food