Message-Id: <199808050132.VAA26532@jekyll.piermont.com>
To: drums@cs.utk.edu
Subject: What is Consensus?
In-reply-to: Your message of "05 Aug 1998 00:26:50 -0000."
             <19980805002650.18149.qmail@cr.yp.to> 
Reply-To: perry@piermont.com
Mime-Version: 1.0 (generated by tm-edit 7.108)
Content-Type: text/plain; charset=US-ASCII
Date: Tue, 04 Aug 1998 21:32:53 -0400
From: "Perry E. Metzger" <perry@piermont.com>


"D. J. Bernstein" writes:
> Chris Newman writes:
> > Therefore I will 
> > place it on the agenda for the next IETF meeting where it is easiest to
> > measure rough concensus.
> 
> All of the people who supported the proposal will be at the meeting.
> 
> Most of the people who objected won't be able to attend the meeting.
> 
> Your ``consensus'' will therefore be biased in favor of the proposal.

The IETF assumes that the people attending are sufficient to gauge
consensus. If so few people are against a proposal that the attendance
or absence of one or two individuals makes a substantial difference in
how many people oppose something, then it is likely that only a
handful are in opposition and that the rough consensus is clearly in
favor of the proposal.

The IETF does not ask for unanimity. You can never get a group of 30
people to absolutely agree on anything. The IETF asks only for _rough_
consensus.

If you honestly feel that the only people who are supporting you are
one or two people on this mailing list who cannot attend the meeting,
I believe you have essentially conceded the chairman's judgment that
consensus is strongly against you.

Please note that the IETF is not always "right" in the sense of
technical perfection. We do not attempt to achieve "perfection". We
attempt to produce reasonable work that has decent engineering quality
and is likely to meet most of the needs of most of our users well into
the future. "Perfection" isn't achievable anyway -- there isn't even
any way to measure it.

By most reasonable measures, the DRUMS working group has formed
consensus on a more than "good enough" standard that comes as close to
achieving our goals as is likely to be possible. As I noted, there
will almost always be be some debate on some small points. However,
were the standard to be adopted as presented, mail would continue to
flow pretty damn well in the world, which is really all we need to ask
for. The goal is, after all, to see mail continue to flow well, and
the document certainly does not pose an obstacle to that.

To summarize: perfection isn't an option. Perfection is not
measurable. What we have will work very well, even if it has some
small flaws -- there is no "perfect" standard anyway. The working
group has more or less formed consensus, and the world will not be
harmed by moving forward as the Chair suggests.

Perry