/20's for the needy
At 12:06 02-07-97 -0700, you wrote:
>On Jul 1, Stephen Sprunk <spsprunk at paranet.com> wrote:
>> I'm not after a separate-but-equal net; there are technical problems that
>> creating said PI space would solve. If done correctly, it would:
> The important thing up there, though, is "if done correctly."
> How are we, the networking community, going to ensure that this
> /is/ done correctly? There are really only two possibilities
> that I can think of (though, of course, they aren't mutually
> exclusive): education and punishment.
My expectation was that there would be guidelines about who qualifies for
an allocation out of the block, and rules about future allocations if you
get one; here's what has been bouncing around in my head:
. Must have an ASN
. Must have no more than 4096 PI IPs already
. Must not qualify for a /19 (or shorter) under RFC 2050
. Must be capable of advertising the /20 to peers within 30 days
. Must be renewed every 12 months or will be reclaimed automatically
. Will not be allocated other PI blocks until the /20 is scheduled for return
> By education I'm talking about things like Justin's post to
> NANOG the other day, where he was inviting people to volunteer
> to help the folks who, according to Tony's CIDR report, are
> having aggregation problems (or making mistakes.)
I went through the CIDR report once and emailed the contact for every AS in
the "most wanted" list, plus all of the "potentially interesting
aggregates" section, to offer any assistance they needed to make their
advertisements more efficient; only one provider (ANS) admitted any
problems (they were advertising 96 more-specifics for a B), and the rest
claimed to have everything configured properly and that they didn't want
> And by punishment, we'd have to come up with fair but firm
> policies for revoking allocations or something like that.
I would think that revoking past allocations would be too expensive to
defend in court; denying new allocations should be enough to bully most
ISPs into compliance.
> BTW, I'm not in any way convinced that these are the best or
> even the only ways to make something like this work...I'm
> really just brainstorming here.