[ntp:questions] Re: Why ntpd is losing out to openntp at BSD

David L. Mills mills at udel.edu
Thu Oct 14 23:43:43 UTC 2004


Sure, any machine with any SNTP client I have found can make a few tens 
of milliseconds, as long as the computer clock oscillator is within 100 
PPM or so and the network is well behaved and the NTPv4 extensions are 
not required. I certainly would not want a SNTP client like your to 
become a server and export 50-ms offsets to a population of dependent 
clients. If OpenNTP is well written, documented and tested, I would very 
much like to replace the current SNTP client presently in the NTPv4 
distribution with it. Maybe the ISC guys can make that happen.


Peter Hessler wrote:

> On Thu, 14 Oct 2004 21:55:47 +0100
> david at djwhome.demon.co.uk (David Woolley) wrote:
> :The non-legal, non-security, argument that they use is that most people
> :don't need better than about half a second accuracy (even if the starter
> :of this thread is complaining about 10ms errors).  If you add on the
> :demands for sub-second initial time setting, step free response to major
> :input steps, and cooperating isolated islands (which account for most
> :of the current FAQs), the official ntpd is at risk of losing out to
> :poorer substitutes.
> I use OpenNTPD on my time server, and I am off by 0.027525 seconds at this
> time.  Except for the initial sync, I am never off by more than 0.05
> seconds.

More information about the questions mailing list