In message <[EMAIL PROTECTED]>, "Richard J. Sexton" writes:
>
>>It's also worth noting that virtually every other major Internet
>>service has been swamped by unexpected load. Predicting load, and
>>engineering for it without prior experience in that particular kind (and
>>popularity) of service is just plain hard.
>>
>> --Steve Bellovin
>
>You really think so Steve? The Porsche mailing lists have 34,000 subscribers
>and it seems to me the notion of inviting the world to vote on how the
>Interent will be run (especially in light of all the "outreach" talk
>that's bandied about) will have a much greater auduience than
>a bunch of Porche owners. Anticipating 5000 users just weems wacky;
>if it were me I'd may sure it could work for a million with
>a contingency plan in place if/whe it exceeded that.
I really think that I've seen a lot of services become overloaded, including
ones where the designers were warned to "plan for success".
Is this such a case? I have no idea -- I haven't audited their staff,
their plans, or their budget. Planning for a million users is
*expensive*, especially in a case like this where there's a significant
amount of manual processing. ICANN doesn't have an unlimited budget;
they'd be acting iresponsibly if they did anything but plan for a
rational estimate plus a decent amount of headroom. But even if you
have a perfect estimate of the load, sizing your systems for a given
load isn't easy. I'm definitely not calling this dishonesty; I'm not
even calling it incompetence, since I've seen too many people I respect
fail at similar tasks the first time around. The touchstone for either
issue is what happens the second time around, where you have experience
to draw upon.
To me, the real question that needs to be answered is whether or not
the registration woes are sufficiently bad and have excluded a
sufficient number of people as to warrant postponing the election.
That's a serious step, and would mean, among other things, that the
interim board would continue longer.
--Steve Bellovin
