[arin-ppml] IPv4 Depletion as an ARIN policy concern

William Herrin bill at herrin.us
Sun Nov 1 13:47:13 EST 2009

On Sun, Nov 1, 2009 at 10:27 AM, Lee Howard <spiffnolee at yahoo.com> wrote:
>> Not possible. Not even in 3 years.
> I'd like to dig into that denial some more, preferably without argument
> by toilet analogy. :-)
> Your objections, as I recall (maybe you could list them again, so we
> can discuss what needs to happen on each one):
> 1.  IPv6 transit is unreliable
> 2.  An dual-stacked client with IPv4-only connectivity may try IPv6
> first, and wait for timeout before successfully using IPv6.


I think if I abstract my complaint it'll make a little more sense:

1. Nothing inherent to IPv6's design acts to make it more reliable than IPv4.
2. Given two systems whose theoretical reliability is identical, the
practical difference in reliability will tend to be a function of the
relative experience and investment in each.
3. The investment in and experience with IPv6 is paltry compared to
the investment in IPv4 and the gap is widening, not closing.
4. From 1-3, I expect (and in fact observe) that IPv6's standard of
reliability in virtually all of its aspects is significantly behind
5. Due to a poor architectural decision by the IETF (IPv6 first, fall
back to IPv4), I can't make effective use of IPv6 *at all* unless I
deliberately ignore #4 and choose to accept degraded functionality on
my system.

Arguably the massive deployment of NAT necessary to extend IPv4 will
alter point #1, with the result propagating through the logic chain.
Arguably the address acquisition cost of continuing IPv4 post
depletion will tip the scales so that the reliability/cost ratio tips
in favor of IPv6. Arguably some unusually valuable capability will be
identified in IPv6 that doesn't exist identically in IPv6, altering
the benefit/cost ratio between IPv4 and IPv6. I'm prepared to act, but
I'll believe it when I see it.

#2 can be observed in the details. Tools like NAT and RA-guard which
don't properly exist yet in IPv6. Complications figuring out who will
talk to who and how in the backbone. Etc. Etc. For every detail
cleared up, there are another handful waiting behind it.

If not for #5, IPv6 would have a much easier time getting past the
"worthy enough to deploy" barrier. There's a moderate risk-cost
associated with only being able to use IPv4 given the uncertainty
surrounding the end of the free pool. Weighed against the risk-cost of
crashes, malfunctions and security breaches due to configuration and
software changes to enable IPv6, the costs are almost in parity. But
the fact that the systems so-enabled will attempt IPv6 first and only
fall back to IPv4 just kills the whole equation.

Of course, that pendulum could swing the other way too. When whatever
happens post-depletion settles out into a routine, the risk-cost of
continuing with only IPv4 could go way down.

Bill Herrin

William D. Herrin ................ herrin at dirtside.com  bill at herrin.us
3005 Crane Dr. ...................... Web: <http://bill.herrin.us/>
Falls Church, VA 22042-3004

More information about the ARIN-PPML mailing list