[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [wg-d] Robert's Rules
On Tue, Aug 10, 1999 at 12:32:55PM -0700, Randy Bush wrote:
> > Also, Robert's Rules can be used to slow down a process when somebody does
> > not want anything to happen. Anybody can just ammend aproposal (only needs
> > somebody to second) and ask for a vote. This can go on forever.
>
> not true.
But it can go on a long time...
> javier, we have a process that is hitting a wall and soon to die partially
> because there is widespread perception that the process is sorely broken and
> many people have been treated unfairly and rudely. this opinion is not held
> only by the crazies.
>
> in my culture, the way to fix that is to give the process some rules to
> induce fairness. someone has suggested some widely used rules. do you have
> an alternative to recommend, or shall we drive the car right into the brick
> wall?
See the attached:
Decision Processes in Working Groups
There have been claims that all decisions in DNSO working groups
should be made through "formal" votes, and that the IETF model of
"rough consensus" is inappropriate for the purposes of the DNSO
working group. This brief paper explores this issue.
It is important to note at the outset that the important distinction
is not whether WGs vote or not -- voting happens in IETF WGs, and is
explicitly mentioned in RFC2418. The point at issue is whether it is
to be *required* that *all* decisions in all WGs be put to a vote, on
demand.
History and Background
In fact, the issue of decision processes ("formal voting" and "rough
consensus" are just two examples) was considered at great length
during the DNSO formation process and earlier, and various voting
regimes were explicitly considered. There were several proposals for
WGs, or "research committees" or just "committees", with sometimes
very complex procedural rules. (1)
The final result was closest in spirit to the "Irish Proposal" -- the
"General Assembly" language came directly from the Irish Proposal
(though there were similar ideas in the other drafts). The final
language in the bylaws is very sparse, only saying that the GA is a
"forum" open to anyone, and that it forms a resource for filling "WGs
of the GA" that the NC may empower.
The bylaws also allow for the NC to empower research or drafting
committees. A careful reading reveals that these committees do
not need to come from the GA. WGs, however, are a feature of the GA
-- the phrase "WGs of the GA" is used in parallel with "research and
drafting committees".
The Irish Proposal language, I was told by one of its proponents, was
intended to model the GA on the IETF, and the parallels are obvious.
The intent was that the cost of effective participation in the work
of the GA was to be very low, with most activity occurring through
email lists, with an open consensus based method of developing
policies (2). It was not intended that the GA would be the only
source of policy, and this intention is carried through in the
current ICANN bylaws.
The implication of the language in the bylaws is that WGs of the GA
should be open to all. A corollary is that there should be no upper
limit on size. However, it is important to remember that the NC has
broad leeway and great flexibility in how it parcels out work, and
could, for example, give a matter to a drafting committee composed
entirely of members of the NC, with no members otherwise drawn from
the GA. It could also create a committee of fixed size and
restricted membership, drawn from the GA.
That is, the nature of the relationship between the NC and the GA is
such that WGs could reasonably adopt a wide variety of forms, and it
would therefore not be productive to try to set rules that too
strictly control the structures in advance. Instead, a better model
would be one where flexibility is allowed. On this basis alone it
would be unwise to mandate a particular decision process.
Issues with Voting
Certain issues concerning online voting have come up repeatedly.
Issue: voter credentials
The level of openness implied in WGs of the GA pose certain problems,
which can be illustrated by the following message to another list,
from Randy Bush (used with permission):
> at this point in the group's procedural maturity, i think the best we can
> do is consensus. my personal definition of consensus would be no
> substantial (by that i mean substantiated by hard arguments and/or facts)
> objections from folk not in my procmail filter (i.e. i am sure the usual
> destructive sickos will scream but i really don't care).
Randy refers to "destructive sickos", but in fact there are other
categories of problem voters. I sent around a message earlier
describing other categories of individuals who can be destructive to
a voting process -- individuals who are brought in solely to stuff
the vote, basically. Given an open membership (3), such behavior is
very difficult to control.
We also have to worry about fake voters, ringers, and participants
who are only brought in to vote, but otherwise do not participate and
have no interest in the proceedings. (4)
Issue: Efficiency of voting
The "legislative history" of DNSO WGs indicates that they are
intended to function primarily over email, even more than WGs of the
IETF. Given global participation, the bare minimum of time for a
vote would be on the order of 24 hours, and experience has shown that
24 hours is not really sufficient for a fair vote. Votes so far in
DNSO WGs seem to be on the order of 3 days, and there have been
notable problems with people not being available, both as candidates,
and in terms of missing votes because of travel etc. Voting periods
in PAB were one or two weeks. (5)
Formal votes, on-demand for any issue, at three days per vote, would
seriously slow down any process, and for all practical purposes is
unworkable, because it provides a mechanism for infinite slowdown of
the proceedings. This is not to say that occasional votes, on large
matters, might not be very useful. The PAB charter
(http://songbird.com/pab/draft-charter.html) describes a process with
a hybrid of consensus and voting on large issues. (6)
Issue: non-representative character of WGs
WGs are fundamentally cannot be considered representative entities.
They are self-selected small groups, and thus they are intrinsically
biased and non-representative. This claim caused some controversy
earlier, but it is almost completely self-evident:
Consider WG-C, which is deciding, presumably, about the addition of
new gTLDs to the root zone. There are 68 members of WG-C. On the
other hand, the true stake-holders, the entities that will be affected
by decisions concerning the addition of new gTLDs, number in the
millions, most of whom have never heard of the DNSO. There is no
meaningful political sense in which those 68 members can be said to
represent the stake-holders in the decision.
Likewise, consider WG-A, with 15 members. It is obvious to most
people that it could not, not in any meaningful sense of the term, be
said to be representative of all the stake-holders who would be
affected by their decisions. However, this would have been true,
even if the entire GA had participated. (7)
This non-representative characteristic of WGs (or "research/drafting
committees", as well) was explicitly considered in the deliberations
surrounding the creation of the DNSO. It was realized that the down
side of free, fluid, self-selecting WGs was that it was perfectly
possible that a WG would produce a result that was completely
unacceptable to everyone else. That is why the NC can reject the
conclusions of a WG, either by simply returning the matter to the WG,
or chartering some new body to decide what to do about the issue.
Issue: Excess Formality
The enthusiasm of even the most ardent supporters of Robert's Rules
seems to have cooled somewhat, and they are now trying to sell
informal versions, by saying that all that is really involved is the
chair calling for votes at "decision points". This is, however, a
gross oversimplification -- "Robert's Rules of Order Newly Revised"
is 700 pages long, and there is a reason you need a parliamentarian
when you use them -- while the basic idea may be simple, in
implementation they are rather complex.
But more seriously, the full rigor of Robert's Rules doesn't
translate well to the email environment of a WG. The time delays
inherent in email have already been mentioned. But there are also
problems that come not just from the characteristics of email, but
from the characteristic of openness -- I'm not sure how Roberts rules
would deal with a situation when voters could arrive and leave at
will. There are also problems of visibility -- when someone leaves a
physical meeting, it is obvious, but when someone stops paying
attention to their email there is no external cue. Another incongruity
that comes to mind: in a physical discussion someone speaking for an hour
would be severely disruptive, but when someone sends a 20 page
missive to an email list, it just gets ignored. Yet another
problem: the matter of unreliability -- while email is generally
very reliable, on occasion it does fail. It's not clear how Roberts
rules deal with the gods randomly striking the participants dumb.
How does one call a meeting to order? How do you rule someone out
of order?
One can multiply these examples almost at will. The fact is that
Robert's rules really are intended for a different environment, and
it would take a fair amount of tweaking to modify them to fit email
appropriately, if it can be done at all.
We can contrast that with the IETF WG rules, which, of course, have
been tuned to the email environment over quite a few years.
The proposal now seems to be that we should follow voting rules
*loosely modeled* on Roberts Rules. In that case, I would suggest
that we simply follow the PAB rules, since they fit this bill.
Observations and Conclusions
Here are some, in no particular order:
1) People are making more out of WGs than they should. WGs should
be a flexible means of doing work; people are fixated on procedural
matters, and strict formalisms. Formalisms should be available for
use, but they may very well not be appropriate for every WG.
2) There is widespread misunderstanding about the role of the GA.
The GA has great flexibility to self-organize, but its role in the
DNSO is rather constrained, and it fundamentally has *influence*, not
direct power. While it may appeal to some to turn things around and
make the GA the basic power base in the DNSO, that is not the way the
bylaws are written, and it is not part of the charter of this WG to
change them.
3) Following Roberts rules strictly is impossible. An ad hoc voting
system based on Roberts rules is possible, but the issues of debating
over email must be addressed. If the goal is to get a voting system
into an online debate, then the PAB rules for debate and voting are
pretty well thought out.
4) Personally, I think that the GA should be organized in such a way
as to have representation on the NC. But that requires a certain
amount of self-organization to take place, and should be the topic
of another forum.
5) The nature of online discussion require human moderation, even
more so than with face to face discussion. The attempt to
substitute formal procedure for human moderation is both unwise and
unnecessary -- a WG can disband, or find another chair, if the
situation is intolerable.
6) Online votes take far more time than a show of hands. When you
add the fact that we really have global representation, you have to
allow at least a few days for a vote. Therefore, you must have some
way to limit the number of votes taken.
7) At the simplest level, votes are an information gathering
mechanism. To turn them into decision procedures you have to apply
decision rules (such as "the proposition with the most votes wins").
The decision rules are just as important as the voting procedure
(though they are often intertwined).
PAB Rules of Order
Here are the PAB debate/voting rules, as modified for a GA WG. Note
that there are various parameters for times and percentages -- these are
obviously all tunable. All interactions mentioned take place on the mailing
list:
The WG shall arrive at recommendations, advice, resolutions,
internal policies, and other common positions by rough consensus
methods, as follows:
1. The Chair may attempt to determine a consensus concerning a
given issue by announcing a Call for Consensus to the WG list.
The Chair must announce a Call for Consensus on a given issue if
10% or more WG members request it of the Chair, through messages to
the list.
2. After a reasonable period of debate, generally not less than
one week and not more than one month, the Chair shall announce
either that a preliminary consensus has been reached or that it
has not been reached.
3. If the Chair announces that a preliminary consensus has been
reached, then following the Chairs announcement, a reasonable
period, generally not less than one week, shall pass in order to
allow dissenting views to be heard.
1. If 10% or more of the Active WG Membership dissents from
the Chairs announcement of a preliminary consensus, the Chair
shall have the option of either tabling the proposition or
calling for an election on the preliminary consensus by
informing the WG list.
2. If less than 10% of the Active WG Membership dissents from
the Chairs announcement of a preliminary consensus, the Chair
shall announce that a final consensus has been reached, and the
Chair shall, if appropriate, communicate the results to the
NC, either through the NC Representatives to WG or directly.
4. If the Chair announces that a preliminary consensus has not
been reached, and the Chair does not decide to seek an election
on the matter, it may continue to be debated on the WG list, but
the Chair may not be compelled to seek another consensus on the
matter for three months. But if the Chair decides that with
small modification a consensus could be reached on the matter,
the Chair has an affirmative duty to find the correct compromise
that will allow a consensus to be reached.
5. In the event that the Chair announces that a preliminary
consensus has not been reached, the Chair may determine, at the
Chairs discretion, that the matter is important or timely enough
that a definitive position by the WG is required. Examples of
such important or timely matters might be: a request from the NC
for a recommendation, or a matter that is of clear and pressing
importance. In such cases, the Chair may call for an election on
the matter. The election shall be carried out via the WG using
the normal voting mechanism that is used for the elections
of officers, and notice shall be given to all Active WG Members
at least one week before the election takes place.
1. If 67% or more of the Voting WG Members are in favor of
the proposition being voted upon, the proposition shall have
passed, and the Chair shall announce that a Consensus has been
reached; otherwise, the proposition shall have failed, and the
Chair shall announce that a Consensus has not been reached.
2. In the event of an election, each WG Member shall have one
vote.
Note that these rules do not deal with the issue of a fluid
membership, nor with credential problems. However, those are *hard*
problems....Note that the basic mechanism remains a consensus
process, with an option to force a vote, just as Karl describes, but
also with a way to limit votes if it is necessary.
================================================================
(1) In all, there were at least 9 distinct attempts to come up with
bylaws for the DNSO. The INTA draft, the Barcelona-Monterrey draft,
the AIP draft, and the Paris draft, as well as several intermediate
compromise drafts, all had different proposals for committee or other
similar structures. The drafts borrowed a great deal from each
other, and there was a great deal of overlap.
(2) In contrast, for example, to the INTA draft, for example, which
had most operations taking place through face to face meetings.
(3) Another solution is to restrict membership in WGs. There
are difficulties if you go down this path.
(4) In contentious WGs we might well get a voting "arms race" where
people send email to everybody they know to gather armies of voters.
This behavior is indistinguishable from "notifying legitimate
stake-holders", and so cannot be controlled. But there is no question
that such behavior would be very destructive to any hope of forward
progress in a WG.
(5) The Policy Advisory Body (PAB) of the gTLD-MoU was in fact what
some would like the WGs to be -- a large deliberative body that ran
used voting. See http://songbird.com/pab, and especially
http://songbird.com/pab/voting.html. However, every member of PAB
had credentials, because their organization signed the gTLD-MoU.
Even in that case, however, there was an attempt to completely take
over the organization -- see http://songbird.com/pab/takeover.html.
(6) The fluid character of the electorate in a GA WG would make
application of the PAB model somewhat problematic, though there are
some elements that could certainly be used. The success of the PAB
model (and it was quite successful) depended in part on the fact
that the members were validated through the registration of the
signatories at the ITU. GA WGs have no such control.
(7) There is a further issue with WG-A: it's 15 members are judging
a far more extensive worldwide process that by any reasonable
objective measure has far more input than WG-As 15 members could
ever hope to achieve.
--
Kent Crispin "Do good, and you'll be
kent@songbird.com lonesome." -- Mark Twain