rough consensus (was Re: "trouble maker")

Harald Tveit Alvestrand harald at alvestrand.no
Thu Jun 26 01:45:45 CEST 2003


Note: There is not consensus in the DNSEXT WG that Phill's description of 
the events of that group is the right one. In particular, as far as I can 
tell, the WG Chairs and the responsible AD disagree with him, but they are 
not the only ones.

This can be a problem when using examples (which I am very much in favour 
of, btw!) - all I'm 100% sure of is that the DNSEXT WG saw several voices 
raised on the issue Phill is describing, and that they did not come to 
consensus.

The chairs reported that lack of consensus as "insufficient consensus for 
change", and concluded that the WG had to ship the documents in question 
without the proposed change.

Text of announcement below.

                    Harald
-------------------------------------------------------------------------

the wg last call on [ds-only] opt-in is over.  it's a tough issue
and the wg discussion has been muddy, to be kind.

1 - there seems to be consensus that the spec is complete and
    correct.  this was determined in a separate last call.

2 - is there consensus that opt-in adds complexity?  probably yes,
    but folk disagree on how much and disagree on the impact of
    that complexity on dns operation and on security.

3 - is there consensus that opt-in changes the security model from
    'simple dnssec'?  yes.  is that change bad?  opinions differ.

4 - is there consensus that deployment of dnssec is important?
    yes.

5 - is there consensus that opt-in may help deployment of dnssec?
    yes, for a few very large zones.

6 - it might not be obvious how to come to a decision here - or
    even whether we can come to a decision.

    the question asked of the WG was:

    "is there consensus in the wg, to allow ds-only opt-in in
    zones?"

    the way this is phrased the default action, which is what
    applies should there not be rough consensus, is to not add the
    opt-in feature to the protocol.

    one might argue that the question was asked in a biased manner
    and the question should instead have been asked as:

    "should we allow or not allow opt-in zones?"

    first, it is traditional and reasonable to place the onus of
    consensus on *additions* to protocols, since it is likely to
    lead to less features in protocols if each added feature needs
    wg rough consensus.  i.e., this was normal wg procedure.

    second, and possibly more important, had the question been
    asked that way, we would have the equivalent of a hung jury
    with no process for how to move forward at all.

7 - is there consensus that consensus-5 should override consensus-2
    and consensus-3?  discussion varied, folk had opinions on both
    sides.  underneath the flamage, there was actual disagreement,
    with knowledgeable parties on both sides of the fence.  this is
    not consensus.

given that we all want to move forward the chairs, given the
process, have no choice but to come to the following conclusion:

    there is no consensus to add opt-in.

randy & olafur






More information about the Problem-statement mailing list