root cause issues in working group management

[Fred Baker]

At 06:14 PM 12/11/2002 -0500, Theodore Ts'o wrote:
>What problem are we trying to solve here?

Remailing of a private note I sent earlier today. I think the fundamental 
problem relates to "too much heat and not enough light".
-------------------------------------------------------------------------
At 09:56 AM 12/11/2002 -0800, someone wrote:
>Limiting the number of working groups hopes that the number is the right 
>number and that it leaves the cognizant area director enough time to give 
>proper wg focus. We have no empirical basis for choosing the right number. 
>We have no empirical basis for believing that choosing the right number 
>will produce the desired effect.

This is pretty much my concern with the proposals, both with limiting the 
number of WGs and limiting their duration.

Personally, I think the issue in this part is not that the calendars are 
too long, but that they are ignored. The milestones (as I frequently 
complained to the IESG during my tenure on it) are not treated as 
management-by-objective steps. There is no penalty for a working group 
ignoring them entirely and doing something different (as happened in RAP) 
other than that the working group chair will probably never be trusted with 
a WG again. I think the duration of a working group should be as specified 
in the list of milestones in the charter, and that needs to be agreed and 
maintained between the AD and the WG as represented by its chair, and 
actively managed to. If the milestones go seriously by the board, that is 
an indication that the WG is moribund or misdirected; its work needs at 
minimum to be replanned and perhaps put the WG out of its misery.

The behavior of an AD I once worked under comes to mind. A WG would get a 
hard date set, and after that date it didn't meet. I didn't agree with the 
policy as stated at the time and I still think the absolute cliff is over 
the top (one or two additional meetings make sense if there is a timeline 
to complete the work agreed to), but the general idea was pretty effective 
as I recall.

I also think we need to look hard at the number of active participants in a 
WG. Every WG I work in seems to be an active discussion among 5-10 people 
with 10-20 interested-and-useful onlookers, plus multiple hundreds of other 
folks occasionally tuned in and otherwise mostly contributing body heat. We 
have tried a number of approaches to addressing this, and I don't know that 
we have succeeded, but the goal needs to be to include that basic few tens 
of people and allow the n*100 to monitor progress and discussion in another 
way.

Let me take a specific example. Diffserv couldn't meet without a room that 
sat 300 people, and on at least a few occasions required a room for 500. I 
just did an analysis of my copy of the mail file (which probably doesn't 
contain all the messages; I knocked out the spam and the rfc-editor mail 
for sure, and probably deleted some other mail, so this isn't a perfect 
test). But I analyzed 3741 messages sent to one working group's mailer over 
a period of perhaps five years. In that time, a total of 410 people posted 
to the list. 347 sent nine or less messages (154 sent a single message), 58 
sent 10..99 messages, and a grand total of five got into triple digits. The 
30 most frequent posters were:

         430     brian@hursley.ibm.com
         335     fred@cisco.com
         160     nichols@baynetworks.com
         145     jh@lohi.eng.telia.fi
         131     smd@ebone.net
         89      binder@3com.com
         79      vaananen@nasbpd01bs.ntc.nokia.com
         71      yoramb@microsoft.com
         66      ferguson@cisco.com
         64      zhwang@dnrc.bell
         57      curtis@brookfield.ans.net
         57      jkr@min.ascend.com
         56      acasati@lucent.com
         56      slblake@raleigh.ibm.com
         53      swb@newbridge.com
         52      mo@uu.net
         50      dovrolis@hertz.ece.wisc.edu
         47      mac@redcape.com
         43      wweiss@lucent.com
         37      andrew@extremenetworks.com
         37      jh@telia.fi
         37      tli@juniper.net
         36      d.black@opengroup.org
         31      james.roberts@cnet.francetelecom.fr
         31      kasten@argon.com
         30      shivkuma@ecse.rpi.edu
         29      kmn@cisco.com
         29      sob@harvard.edu
         29      bound@zk3.dec.com
         28      minshall@fiberlane.com

Not that the single posters were not valued: Itojun, for example, was asked 
a specific question about his IPv6 Flow Label draft and answered it with a 
single message. Also, some of those are clearly the same poster at two 
addresses: kmn@cisco.com and knichols@baynetworks.com are the same person, 
as are andrew@extremenetworks.com and smith@acm.org. But I think the 
example illustrates the point. Most of the work was done by a small subset 
of the folks.

I think if we could somehow identify the "active participants" and the 
"interested parties", and move the n*100 out of the room or virtual room 
without actually sacrificing openness or wide review, and at the same time 
manage by objectives, it would both make IETF meetings cheaper and help the 
work get done more rapidly. I tend to think the key point is solving that 
difficult problem, not throwing experimental procedural band-aids into the 
pot.