--On Saturday, July 06, 2013 14:53 -0700 NomCom Chair 2013
<nomcom-chair-2013(_at_)ietf(_dot_)org> wrote:
I am pleased to announce that we have 140 qualified
individuals who have generously volunteered to serve as
voting members of this year's Nomcom.
Allison,
Given my previous comment about statistical assertions, a quick
look at this list sheds some additional light on cross-sections
of the community.
(1) While you have 140 people from whom to do a random draw, the
"no more than two volunteers with the same primary affiliation"
(RFC 3777, Section 4, Bullet17) rule means that the actual nomcom
pool is only 81 even though some of the people are indeterminate
(I've made some assumptions about company boundaries -- other
assumptions would yield very slightly different results). That
number is obtained by listing the companies and then counting
any company with more that two volunteers as having only two.
Neither the 140 number nor the 81 number is incorrect, they just
measure different things.
(2) Four companies account for 44.3% of the volunteers. As
others have pointed out, while having a very large number of
volunteers cannot produce more than two Nomcom voting seats due
to that restriction, it can virtually guarantee (statistical
randomness notwithstanding) that one will end up with one or two
seats. Specifically, if the people selected constitute a
cross-section of the volunteers, then more than 10% of the pool
(14 volunteers) would predict to at least one seat. Two
companies exceed that number and a third is very close.
Randomness could make the actual numbers either better or worse,
but, if an organization's goal were to assure at least one seat,
that seems easily accomplished. Of course, having lots of
volunteers doesn't imply that an organization was trying to
accomplish any such thing -- it could just have a lot of
public-spirited IETF participants and policies that allow people
to commit the time.
(3) It is probably too late to even discuss it for this year
(see below) but it occurs to me that, if one wanted to minimize
the odds of organizations trying to game the nomcom selection
process, it would be rational to do a two step draw, first
randomly selecting two volunteers from any organization offering
more than two and then including only those two in the final
selection process. On the other hand, that would give you
around 81 candidates for the final selection this year. If
running the final selection against order 140 people rather than
order 81 causes the community to believe that it has a better
sample, then that option probably would not be appropriate. I
am not, however, convinced that we would actually have consensus
for minimizing those odds, nor about whether a company's ability
to nearly guarantee that at least one of its employees will be
on the Nomcom by providing a large fraction of the volunteers
violates the provision of Section 4, bullet 16, of RFC 3777
requiring "...unbiased if no one can influence its outcome in
favor of a specific outcome".
Actually, if I read Bullet 16 correctly, the choice of methods
is up to you, with RFC 2777 (and 3797) being just examples. So,
in theory, you could make that change. I wouldn't personally
recommend it, but you are the one who has personal
responsibility for things being fair and unbiased while this is
just a statistical exercise for me.
best,
john