Date: 15 Feb 1999 05:42:14 -0000 Message-ID: <19990215054214.9088.qmail@cr.yp.to> From: "D. J. Bernstein" To: ietf@ietf.org Cc: drums@cs.utk.edu, iesg@ietf.org Subject: DRUMS procedural problems: a case study References: <19990213023232.28291.qmail@cr.yp.to> <2948.918973759@munnari.OZ.AU> <19990214111530.3897.qmail@cr.yp.to> <4.2.0.24.19990214233936.00a709f0@shell2.bayarea.net> Dave Crocker writes: > Again, Dan, you misunderstand the process. I understand the process just fine. The document editor makes whatever changes he wants to make. In cases where the rest of us agree with the editor, there's no problem. In cases where the rest of us disagree with the editor, the burden is on us to (1) continue raising objections until the DRUMS management agrees to allow the issue at an in-person meeting, (2) prove consensus at the meeting, and (3) object again when the editor ignores the consensus. In cases where some people agree with the editor and some don't, the editor wins by default. His changes are then labelled ``consensus.'' Anyone who disagrees is challenged to prove consensus to the contrary. Here's one of many examples. RFC 821 says that an SMTP client ``should not'' close a connection without waiting for a reply to QUIT. John Klensin changed ``should not'' to ``MUST NOT intentionally.'' As you read the rest of this story, please keep in mind that DRUMS rejected Klensin's change at the August 1998 DRUMS meeting. According to the minutes: Consensus is to change "MUST wait" to "SHOULD wait" (and some would prefer "MAY wait" or nothing at all). Back in July 1996, I suggested a careful transition away from QUIT. Four hours later, after a few people had expressed interest, Keith Moore said It's not worth the trouble to write it up, or even to argue about it. I believe we probably have rough consensus that QUIT should stay. Are there any objections other than from Dan? When I reminded Moore that he had misrepresented DRUMS consensus before, he said I'm reasonably confident that this will be the consensus. ... I don't want to see us waste time discussing yet another issue for which there isn't any significant support for changing the status quo. After more people expressed interest, Chris Newman tried to cut the discussion short by calling for ``concensus'' on his own minority views. He subsequently apologized. In August 1997 I announced that I was no longer going to wait for a QUIT response in qmail. (Exim now behaves the same way; I haven't checked other MTAs.) Two people responded with personal attacks. In June 1998 I again objected to Klensin's ``MUST NOT.'' Nick Shelness said Dan has been beating this dead horse for quite a while and seems to still be in a minority of one. I responded: ``This is not the first time that the QUIT supporters, who I suspect are a tiny minority among implementors, have claimed that their position is the DRUMS consensus, and attempted to discourage further discussion. Their behavior will make it easy for implementors to discredit DRUMS.'' At this point Newman tried once again to close the discussion: I certainly don't find the QUIT issue compelling enough to delay the publication of a revised SMTP spec and extend the DRUMS working group for a few more months. Unless others do feel it is that compelling I will consider the issue closed. Need I point out that the issue was raised in 1996? In any legitimate standards organization it would have been settled at that time; there would have been no need to point it out again two years later. After more people spoke up in favor of my position, Newman tried _again_ to close the discussion: If there are fewer than two objections to any of the proposed resolutions here, I will assume rough concensus has been achieved. QUIT was item #22; the proposed resolution was to leave Klensin's ``MUST NOT'' in place. Newman didn't care that there had _already_ been more than two objections; as usual, an extra burden was placed on people who disagreed with the document editor. After this attempt failed, Newman finally put ``reduce to SHOULD NOT'' on the list of issues to be discussed at the August 1998 meeting. The result, as quoted above, was consensus against Klensin's change. Klensin still didn't fix his document. (From his editing notes it would seem that he somehow managed to misinterpret the meeting minutes.) I caught this in January 1999, during the smtpupd Last Call. Before Last Call, Newman had stated a policy formulated at the December 1998 IETF meeting: he would assume consensus against any objection that wasn't stated by four people. But what does this mean when the document was changed by the editor without authorization from the working group? I posed a question for Newman: ``In cases like this, where DRUMS has _already_ agreed that the document should be changed, are you still demanding that four people speak up?'' For reasons that still haven't been explained, Newman didn't respond. After the Last Call period, Newman asked someone to prepare a summary of the discussion, excluding objections that weren't raised by four people. He hasn't stated that there will be a new Last Call period. I complained to the IESG a week ago; Newman _still_ hasn't answered my question. Is Klensin going to fix the document? It probably didn't take Klensin more than a minute to make his change. I've spent hours trying to get him to fix it---even though I'm in the majority! The document editor's ideas aren't competing fairly with the ideas of everyone else in the group. ---Dan