[arin-ppml] Did 2008-6 provide what Board needed?

Tue Apr 14 12:09:46 EDT 2009

In a message written on Tue, Apr 14, 2009 at 08:49:08AM -0500, David Farmer wrote:
> The Board's solution is to implement immediately and eliminate the sunset 
> clause.  This definitely clears up the unpredictability and uncertainty, and 
> makes network planning easier.  However, this solution is inconsistent with 

Actually, I think this just traded one form of unpredictability for
another.  When this policy was passed I think most people expected
it "near" IANA run-out.  We can argue a lot about what that means,
but even those with the most aggressive predictions have told me
9-12 months.

However, all the signals the board sent were for immediate
implementation...

https://www.arin.net/about_us/ac/ac2009_0219.html

   John stated the policy will already be in effect when it goes
   before the community at the ARIN Public Policy Meeting in April,
   in accordance with the 'Emergency Special Policy Actions' outlined
   in the ARIN PDP.

That's really just another form of unpredictability, if the goal
was to have people plan ahead then springing a policy change on
them does not allow that to happen either.  I don't think anyone
expected the Board to both take emergency action and implement it
ASAP, no one had in their plans that next month they might be trading
in IP addresses.

In hindsight, I think we both got it wrong.  I think triggering on
IANA exhaustion is likely a little too late for folks to properly
prepare, and the AC and community should take our knocks for that.
I think the Board's swift action was simply the pendulum on the far
opposite side, catching people off guard just as much.

> I believe that we can solve the unpredictability and uncertainty while 
> remaining mostly consistent with the compromise within 2008-6.  I believe 
> the key to solving the unpredictability and uncertainty is to provide a specific 
> start and end date for IPv4 transfers to Specified Recipients. Then provide a 
> safety trigger to start early if IANA exhaustion occurs significantly faster than 
> anyone expects and an automatic extension if IANA exhaustion takes 
> significantly longer than anyone expects.

I'm not necessarily opposed to hard dates, but I feel they make a
lot more sense for the sunset than they do for the initial
implementation.  There has been a lot of speculation about a "run
on the bank", or about particular players who could "request all
remaining IPv4 space".  We would look pretty silly if we picked
what appeared to be a safe date now, and then one of the corner
cases occurred and the Board had to take emergency action again.
Indeed, I think in part the idea of implementing it now was so the
Board was sure there wouldn't be Emergency Round 2.

In the LA meeting, the number report
(https://www.arin.net/participate/meetings/reports/ARIN_XXII/PDF/wednesday/inr_status_report.pdf)
shows approximately 12 /8's were issued in 2007, which was the
highest year yet (note 2008 is incomplete in this chart due to the
meeting time).

My proposal then would be a trigger date of 17 /8's, the 5 in the
everyone gets one global proposal, plus a "years supply" at the
highest rate we have seen.  For reference we are at 32 available
now.

If all goes well that splits the remaining time in the middle.
People have plenty of time to prepare for a world with transfers,
and it should kick in plenty far before exhaustion.  Should some
wacky corner case kick in where someone shows up and justifies 40
/8's one day it still automatically triggers, keeping the Board
from having to take more Emergency action.

Regarding the sunset, I think it matters a whole lot less.  If
people can prepare for transfers to start in approximately 2 years
they can prepare for them to end in the same time frame.  I think
sunsets of 3-5 years from when transfers start is perfectly acceptable
and provide lots of warning of when the date will be.  I have a
whole lot of trouble reconciling that it's ok to start the transfers
"right now" with no notice, but 3 years to plan for them to possibly
end is not enough time.

The whole problem with the sunset is if we start too early, the
policy ends too soon after IANA exhaustion.  It has nothing to do
with planning or certainty.  To that end I would recommend a sunset
of 3-5 years after the date IANA hands out the last /8.  That should
make the total policy time 4-6 years (given what I proposed above
for an implementation time).  It's extremely rare to see serious,
detailed planning more than 5 years out so that would seem to be
plenty of time.

And, of course, as plenty of people have pointed out, policy can
always be changed.  If a 3 year sunset isn't good enough we can
change it later.  Yes, it's the same argument folks use for why we
don't need a sunset, we can always repeal the policy later.
Both are true, it's all about what signals you want to send...

-- 
       Leo Bicknell - bicknell at ufp.org - CCIE 3440
        PGP keys at http://www.ufp.org/~bicknell/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 187 bytes
Desc: not available
URL: <https://lists.arin.net/pipermail/arin-ppml/attachments/20090414/d5aa1c4a/attachment.sig>