[ntp:questions] How is the NTP build tested?

Harlan Stenn stenn at ntp.org
Mon Jan 13 21:34:09 UTC 2014

David Taylor writes:
> On 13/01/2014 14:34, David Woolley wrote:
> > On 13/01/14 10:22, romain.pontida at gmail.com wrote:
> >
> >> againstthe reference implementation, but how do you prove that
> >  > the reference implementation is correct?
> >
> > I thought the reference implementation was correct by definition.
> Ideally, you need a version built purely to the paper specification by
> a "clean-room" team, and then see where the differences lie under
> testing.  I can't see that happening with the effort currently
> available, to be honest.

I'd love to see it, and if this is important enough for folks there
needs to be a funding stream to support that effort.  But I don't really
see the value in that, as we have a very liberally open-source licensed
version of the code and what other benefit would there be to a separate
clean-room implementation?

As to the larger question of "is the reference implementation correct",
I'll point out that there were on the order of 1,000,000,000,000 hours
of NTP operations in the past year alone, and we try to make it easy for
folks to report problems or talk to us.


More information about the questions mailing list