ietf-nntp INN vs the reference implementation

[Brian Kantor]

Russ writes:
>I'd recommend against treating INN as a reference implementation, at least
>until I can come up with some details about how it differs from the
>standard.  INN contains a lot of cruft for a lot of historical reasons
>(and some things that never should have been there in the first place) and
>unless you know INN fairly well, it's often hard to determine whether that
>cruft is useful to document or should just be removed from INN and ignored
>by the rest of the world.

A dissenting view:

We all know that the NNTP specification in RFC977 was made a standard
over the objections of the author (me), which I made because it is flawed
and did not represent current practice even a decade ago.  Practice has
diverged even further from it since then.

INN largely represents the continued development of news transport and
client service.  The community has learned many things from its widespread
use, whereas the "reference implementation" has declined in use; few
important sites use it any more.

Furthermore, during the development of INN, several of the shortcomings
of and ambiguities in the original NNTP specification were resolved by
making INN do what seemed most correct, even when that was in conflict
with RFC977.  Rich Salz and I worked together on some of these points
as he was designing and implementing INN.  We should have documented
those differences and published them as a followup to RFC977, but at the
time, the administrative overhead involved in modifying an adopted standard
didn't seem worth the effort, since everyone would soon be running the
corrected code anyway.

As the current task is to document current/best practice, I would maintain
that those parts of INN that are in daily use represent, to a very large
degree, what current/best practice really is.  To the extent that INN
differs from the standard, we should examine why very carefully, with
a view that INN is more likely to be correct.

After my experiences with RFC977 and other RFCs I have written, I am
extremely wary about any "standard" which is not represented by actual
working production code.  There is a huge difference between a design
document and a working protocol.  We must be careful not to standardize
on paper what isn't working every day in the real world.
	- Brian