Re: [Ntp] I-D Action: draft-ietf-ntp-mode-6-cmds-09.txt

On Wed, Jun 24, 2020 at 5:56 AM Harlan Stenn <stenn@nwtime.org> wrote:
>
> On 6/23/2020 10:15 AM, Salz, Rich wrote:
> >> I believe the WG does not wish to convey "reference
> >> implementation"
> >
> >> The WG as a whole certainly does not believe that.
> >
> > Sorry if I mislead you by not saying "the consensus of the WG"
>
> Thanks, and I believe the mislead was to more than just me.
>
> >> I know some prefer to deny the heritage.
> >
> > I cannot comment to that.
>
> What is your opinion on the subject, then?
>
> >> What other full reference implementation is out there?
> >
> > The IETF does not have the concept of a reference implementation.
> > Maybe it used to.  But for a WG to use the term "reference
> > implementation" it implies a cachet, at least, that is at odds with
> > some of the core principles of interop and RFC's.
>
> The NTP Project's release has long been called the reference
> implementation, in many IETF publications and discussions.  It's not an
> implicit cachet, it's an obvious and long-standing fact.
>
> The NTP Project's code base has, to date, been the base from which all
> previous IETF NTP Standards have been based on.  Many have written
> less-complete implementations that successfully interoperate with the
> reference implementation and with other implementations.

The issue isn't with these historical facts: it's with the attitude
that you've brought to the WG where the reference implementation
decides the future of NTP, and other implementations have to follow
along. This has been a very real problem with assigning extension
codepoints. The contrast to TLS WG to name another I've been involved
with is obvious: Rich Salz and Bodo Moller maintain one of the most
used implementations that has long implemented new drafts ahead of
others. But they've never said "the implementation we maintain will do
X, so everyone else should go along", whereas it seems that at least
twice a month there is some lengthy argument where your position is
that the reference implementation needs more weight than others, for
things like extension space squatting or an unfortunate degree of
bitpacking in extension numbers.

If the future of NTP is to be determined by your implementation, that
raises serious questions about what the rest of us are doing here and
the openness of the process.

> What else could it be called?
>
> Who gains by these attempts to strip the NTP Project's codebase of its
> rightfully- and long-earned title as a (complete) reference
> implementation of NTP?
>
> > The IETF does not
> > distribute source code, therefore it would also be wrong for the IETF
> > to name someone as a reference implementation.
>
> The IETF isn't naming anything.  The name/title came first, from outside
> of the IETF.
>
> By your logic, the IETF is also in no position to deny us that
> name/title, either.

The IETF is a standards body. Usually when someone discusses a
standard and the reference implementation of that standard they imply
the reference implementation has normative power: that matters not
settled from the standard are determined by the reference
implementation. That's not the role ntpd plays for the IETF's time
work, even if past RFCs were descriptions of ntpd's algorithm and
structure.

>
> --
> Harlan Stenn <stenn@nwtime.org>
> http://networktimefoundation.org - be a member!
>
> _______________________________________________
> ntp mailing list
> ntp@ietf.org
> https://www.ietf.org/mailman/listinfo/ntp

--
"Man is born free, but everywhere he is in chains".
--Rousseau.