In message <031117222708.M0103959@mirage.ceres.dti.ne.jp>, you write: 

-> In article <200311171309.OAA0000003481@zel459.zel.kfa-juelich.de>
-> M.Drochner@fz-juelich.de wrote:
->
-> > Is this in-tree ntpd or 4.2.0?
-> 
-> Yes.

Heh, I think that was a multiple-choice question, not a yes/no :-)

-> > 4.2.0 does some strange jumps after a (re)start and takes some
-> > hours to stabilize.
-> 
-> My U5 seems to stabilize with a certain frequency error (-480~500ppm). 
-> 
-> BTW, I've heard it will take much time to stabilize if
-> the initial frequency value recorded in /var/db/ntp.drift
-> is too much different from the real frequency error.

I've certainly experienced this, esp. after the switch to the new timer
code on port-sparc64.  Using the drift file that went with the old timer
code *did* make it take quite a while to stabilize (so much so that I
simply moved the file out of the way and restarted); without any drift
file things seemed to stabilize much faster, and indeed a few hours
later when I looked, the new drift value was vastly different than the
old one.

--rafal

----
Rafal Boni                                                     rafal@pobox.com
  We are all worms.  But I do believe I am a glowworm.  -- Winston Churchill