[ppml] Last Call for Comment: Policy Proposal 2003-4

Peter Bartram peterb at imagine.com
Fri Jun 4 17:30:13 EDT 2004


I've just joined the PPML list, so if I repeat some of what's been said,
that's why.

>"It's probably better to
>wait until there is some industry and consumer pressure to do something.
>The world seems to work better when it's under the gun, instead of trying
>to prepare for something."

I disagree with Mury on this although he makes some other valid points. I
think it's better to prepare for the migration to IPv6 here in the US; so
what follows may seem radical.

I agree with the IPv6 policy change revisions as currently outlined but
would prefer to see a more moderate amount being charged for IPv6
allocations and a scheduled ramp up to get to that reduced pricing level. I
think it would be better to waive the IPv6 fees for three years and then
begin a ramp up in increments of $250.00. IE $0 in 2004-2005-2006 for a
/32, $250 in $2007, $500 in $2008 AND $750 in $2009 and $1000 each year
after that. This way there is an gradually increasing ISP customer base to
justify the IPv6 yearly payments to ARIN. IPv4 addresses will be with us
for some time and their fees plus the new fees should continue to fund the
ARIN operational model.

It is possible to build a ARIN IPv6 conversion financial model. Has anyone
done this? I recommend a 11 year schedule. Why 11 years? To almost fully
replace ARIN's yearly operational costs with yearly IPv6 allocation revenue
requires a financial model with certain basic assumptions.

The first "PUSH" assumption necessary in my mind is this: ARIN's revenue
transition from the current almost 100% IPv4 revenue stream to a blended
IPv4/IPv6 revenue stream has to match ISP allocations and other fees. So
the IAB, ARIN, IETF et all must choose an "ISP recommended" IPv4 allocation
end date to fully stop allocating IPv4 in the Internet and declared it's
date to the world (IE: accomplish this by January 1st 2015 for example). To
clarify:this means to stop allocating blocks, not stop routing IPv4.  This
vision statement of 2015 should provide more than enough time for ISPs and
everyone else to have IPv6 fully deployed in their networks (with Dual
Stack IPv4) and on the way to an all IPv6 Internet. Even if the date is
pushed back as it approaches, the world will have a schedule that it can
try to adhere to. True, politics may intervene. Remember Y2K, now Y15. If
everyone has an allocation end date in mind, this will hasten this
evolutionary process to IPv6 and we will avoid rushes in allocations. If
you have an end date, everyone can plan to convert all aspects of the
Internet in software and hardware, for example, businesses can rewrite
software to work with IPv6 much like they did for the Y2K conversions.
"IPv4 and IPv6 Ships in the Night" routing won't be needed any longer.
Certainly a Y15 discussion should occur. By announcing an end date, the
politics can begin.

What... you might say? Announce the End of IPv4 so we can figure out how
ARIN's revenue model will work? No...what I am saying is that a consensus
of Internet authorities must lead and pick a date and get it in to the
public's mind so we don't end up with the hovering IPv4 exhaustion problem,
BGP routing table exhaustion problems, global network security issues, and
a sudden expensive frantic pace to convert to IPv6 when a major disruptive
technology occurs. ARIN's financial model is just one result of this
decision. If you have an end date, most new ISP deployments will want to
save the conversion cost later and will be more likely to deploy IPv6 in
place of IPv4 or along side IPv4 with dual stack, sooner rather then later.
They will certainly take a closer look at it because of it's eventual
financial impact.

The second assumption is that today the IPv6 address allocation process
shouldn't add significantly to ARIN's operational costs unless a rush
occurs (ARIN already has their current costs in line with operations). This
may or may not be correct. The one thing that will generate a rush in IPv6
requests will be a "killer IPv6 application" like the world-wide-web was
for IPv4. Even if an IPv6 "major app" appears that can generate a large
demand for IPv6 addresses and technology, a far more gradual rollout of
allocations in likely. The impact of very cheap bandwidth everywhere,
Mobile IPv6 with 3G cellphones, laptops, and PDAs, fiber-to-home, and
accelerating the Tier-1 provider cores with products like Cisco's new 90
terabit router with IPv6 hardware routing and Interdomain IPv6 SSM
multicasting will bring this constant change. As more computer platforms
move to 64 bit processors and get less expensive, this conversion probably
won't have to wait the full 11 years.

In  Cisco's December 2003 "Internet Protocol Journal" , Geoff Huston of
Telstra does an excellent job of statistically modeling the end of IPv4 and
puts it out past 2025 at current rates. As he notes though, it only takes
one disruptive technology to change these complex models quickly.  It makes
sense to prepare.

With these two assumptions a PUSH-PULL combined ARIN-RIPE-APNIC etc. model
is needed. The first step is to separate the IPv4 allocation Block billing
from the IPv6 Block billing. Right now they are linked even though IPv6 is
free to current ARIN IPv4 subscription holders. I agree with Huston's
statement that "some would argue that the current situation underprices
IPv4 at the expense of IPv6. No flat rates for new IPv4 blocks three years
in the future.

The second step is The PUSH: IPv4 allocation prices should be scheduled to
rise over the next 11 years on a well advertised schedule with higher fees
being charged for larger blocks and rising in later years as IPv4 address
scarcity increases. The PULL: An ISP's fees should be permitted to go down
as IPv4 allocations are returned and replaced with more abundant, less
expensive IPv6 blocks; while still keeping ARIN in the black. This will
have the effect of getting the bigger providers to convert sooner as they
realize cost savings from the obvious shift to IPv6 allocations and enjoy
IPv6's technological benefits. Or, as we have seen with "smokestack"
industry in the US, some providers will stay with IPv4 and wait until the
last minute or until the "IPv6 Major App" arrives. In either case, once the
majority of larger providers start to shift and competitively promote new
services based on IPv6 into their marketing messages, (such as the end to
end IPv6 security and the overlooked ability in IPv6 of accurately locating
the IP number of sources of spam),  the smaller providers will be PULLED to
follow suit.  (Crossing Geoffrey Moore's Chasm).

>From these above steps, a more precise forecast for the future can be
developed with ARIN's separate IPv4 and IPv6 IP block pricing and
financial/operational model. ARIN and ISPs can plan ahead. And the Internet
can be guided forward with the excellent design of IPv6. Should a rush to
allocate IPv6 develop, revenue from the blended sources of IPv4 and IPv6
can then be finely tuned to promote IPv6 growth over the legacy IPv4 and
fees adjusted up or down to support the cost of ARIN's operations.

RE: comments in previous email:
>> It seems to me that the problem is not on the supply side, but the
>> demand side.  Making it freer isn't going to make anybody want it
>> more.  If I'm wrong, and a few ISPs will say, "The only thing keeping
>> me from deploying IPv6 is the fees I'll see next year" I'd
>> discuss it with the Finance Committee and the Board.  Since IPv6 is
>> waived for IPv4 subscribers, that's a small constituency, though.
>> That is the current policy. IPv6 is free (fee is waived) for new
>> users, and:
>> 	IPv6 fees will not be charged to organizations that are
>> 	current ARIN IPv4 subscription holders.
>> http://www.arin.net/registration/fee_schedule.html#ipv6_allocation
>> It's either free or it's free.
>I understood that to mean they are free until the waiver is up.  In other
>words, when the waiver is up the organization will have to pay for both
>IPv4 and IPv6 blocks.  Am I incorrect?

Finally, I would prefer to see that the billing part of the IPv6 policy be
amended to include the ramp schedule above and to not exclude those that
currently pay no IPv4 fee to ARIN (ISPs and others that have grandfathered
IPv4 blocks, preARIN). Some of us would like to begin with IPv6, or would
like to go direct to IPv6 and can't justify the payment of the implied
$10,000 startup fee (2,500 x 4 years). Developing an IPv6 customer base
will take time and presents a good opportunity for renewed, innovative
growth.  Working with a separate gradually increasing IPv6 yearly fee
structure makes more sense even if that ramp starts in year two instead of
year four for those who are not ARIN IPv4 subscription holders. I believe
the above section of the policy should be amended to read:

          "IPv6 fees will not be charged to organizations that are
            current ARIN IPv4 subscription holders or that hold active
           ARIN IPv4 allocations which are in the IPv4 routing table".

-- Peter

More information about the ARIN-PPML mailing list