On Wed, Aug 17, 2022 at 5:11 PM Christian Huitema <huitema@xxxxxxxxxxx> wrote:
On 8/17/2022 2:13 PM, Fernando Gont wrote:
> Hi,
>
> Can you provide more context regarding what's the issue here?
>
> Is this related to the folks that were not elligible due to being part
> of the ISOC board?
>
> And the objection is against picking "the net folk" in the list, as
> opposed to pick another random number? Or what?
The general issue is about changing the list after the selection has
been done. These changes happen for two reasons: someone is found
ineligible; or, someone declined to participate.
The current process provides for an unpredictable and verifiable random
selection. Given a list of candidates, the randomized process provides
an ordering of the list. The nomcom chair is constrained to select the
first N candidates according to that order. The design accounted for
people found ineligible, e.g., because there were already 2 participants
selected from their company. In that case, the nomcom chair just picks
the next candidate according to the randomized order.
I think that we do not have a big problem if someone is found ineligible
after the selection, and that picking the next candidate in the random
order just works. However, we do have a problem if one of the selected
members decide to decline at the last moment, because at that point the
order of the list is predictable.
So we can see games in which a given
selected member observes that "if I quit, then candidate N+1 is
selected, and I know that my company would prefer this candidate N+1 in
the committee, for example because of seniority." We may be worried
about that kind of games.
Wow. I did not know that there are companies that would consider doing these types of things.
Personally, I have volunteered for Nomcom many times, maybe app. 20, and my number was never selected and I never cared about it neither the company I worked for at the time.
Behcet
The proposal to just run the algorithm again with the remaining
not-yet-selected candidates has the same predictability issue. If the
random seeds are published, anyone can predict who "N+1" will be. To
achieve unpredictability, the process will have to incorporate a new
source of randomness. But that requires some delay, such as waiting for
the next results of a lottery. That may well not be practical. It is
probably not practical at all now, because we are running out of time.
In fact, there are multiple ways to design algorithms like that, maybe
using additional random seeds put in some kind of escrow by the nomcom
chair before the initial lotteries, instead of relying only on external
lotteries. I do not expect than the IETF would converge on a solution in
just a couple of days. Which means, for 2022, Rich has to make some kind
of executive decision.
-- Christian Huitema
