[ntp:questions] Poul-Henning Kamp and re-write of NTP
snews at lordynet.org
Sun Dec 7 20:49:18 UTC 2014
David Woolley wrote:
> On 07/12/14 17:48, Paul wrote:
>> That's unnecessary complexity. refclock_atom is only ~ 200 lines. The
>> just needs to be adjusted so you can build with exactly the driver(s) you
>> need plus PPS without (essentially) forcing all-clocks. There's little
>> reason for a given refclock build to target multiple sources.
> Most ntpd users these days will install binary packages, not build from
I've had pool servers since end of 2009 and was reasonably happy
with stock ntpd that came with NetBSD. Since 2010 I've tried
various refclocks for time from DCF, MSF then GPS. Stock ntpd is
rather old and some options of ntp-dev looked to offer some
advantages so since 2012 I've been building from source then
installing from the binaries on all my systems, now including
four pool servers as from Jan 2013.
Offset of my Sure GPS synced system, not in the pool, is mostly
< 5us, internet servers and systems on the lan are mostly < 250us.
The big improvement I've seen with ntp-dev is that convergence
after a restart is quite fast and reboots of my public servers
often doesn't give any reduction in my pool scores.
More information about the questions