[ppml] Policy Proposal 2005-1: Provider-independent IPv6 Assignments for End Sites - Last Call

[Robert E.Seastrom]

Vince Fuller <vaf at cisco.com> writes:

> The IETF process failed ten+ years ago when a "solution" was picked that had
> been clearly shown to be inadquate at solving the known, hard problems.
> Since then, politics have triumphed over technology and all efforts to
> design a real solution have been effectively surpressed in the name of
> consensus-building around ipv6. There are some people, far brighter than me,
> who continue to try to fight the good fight, but the entrenched interests
> around ipv6 make that difficult-to-hopeless.
>
> (OK, now I'm sure I appear to be a conspiracy theorist, so this will be my
> last posting on this topic)
>
> Without a clear message from the IETF contituency that ipv6 is fundamentally
> flawed to the point of not being worth the time, effort, and expense of
> deployment, there is exactly "zilch" chance that the established path will
> be altered. "Moving forward" down a path that inexorably leads to a crisis
> vindicates those that deny the existance of a problem (think global warming
> here) while the existance of the problem is masked by the early, shallow
> slope of the growth curve.

I see we're in agreement about the process having failed; we disagree
on what to do next.

It's clear to me that the past decade of repeatedly sending the
message that the process has failed has not had the desired effect.
It's not clear to me why you believe that continuing to send the same
message, only more strongly worded, is going to change things.

>> How about a REAL solution from the IETF with a concentration on the
>> "working code" part of "rough consensus and working code" so that we
>> can get on with *repealing* 2005-1 after it's no longer necessary?
>
> Without a commitment to the creation of a solution that obviates 2005-1,
> this is a mistake. If there is one thing we've learned from the development
> and deployment of CIDR, "temporary" solutions aren't, especially when there
> is no clear direction to the "permanent" solution or even a clear
> definition of the problem to be solved.

CIDR wouldn't have happened without people getting _worried_ about
routing table growth.  As long as v6 is moribund, there's nothing to
worry about.

ARIN policy has changed in the past to meet the needs of the
community, real and perceived.  I can assure you that if 2005-1 is
approved there will be one new proposal per cycle to reverse it.  Once
we see widespread deployment of v6 and routing table growth to a point
that is at the same order of magnitude as the current v4 routing table
*and* there is a way to handle the multihoming problem without PI
space, you'll start seeing widespread support for discontinuing or
modifying the v6 PI address space issuance policy, and at that point,
one of the supporters will be me.

Will we be stuck with routing a swamp in perpetuity?  Yup.  That's the
price we unfortunately have to pay for having a protocol that not only
didn't address all problems expected of it when introduced, but has
avoided facing them for a decade.

> The Internet faces pressure for change today: the exhaution of the 32-bit
> IP address space. Spending billions of dollars to deploy a band-aid that
> will only trade this crisis for another one a few years down the road hardly
> seems like a wise investment, especially when one considers that the
> installed base, and thus the cost to deploy a fork-lift "upgrade", will be
> orders of magnitude bigger in the future. Fix ipv6 now and you inconvenience
> a few early adopters who have to roll-out new implementations; fix it later
> (assuming there is a wide-scale deployment and transition from IPv4) and you
> face a second, far bigger, multi-billion-dollar transition. Is that really
> the best plan that the community can conceive? IMHO, disappointing, to put
> it mildly.

I agree that it's disappointing, but find it preferable to the
alternative of not having a migration path away from v4 before we run
out of address space.

I would love to see your "fix ipv6 now" eventuality, but based upon
past history I believe that this is every bit as much a fantasy as the
notion that ipv6 will be a protocol that will last hundreds or
thousands of years.

> In case it isn't obvious, this is a vote *against* adopting 2005-1.

Also in case it's not obvious, I'm in favor of 2005-1.

                                        ---Rob