[arin-ppml] Revised -- Policy Proposal 2009-4: IPv4 Recovery Fund

[Leo Bicknell]

In a message written on Sat, Apr 11, 2009 at 04:01:23PM -0700, Matthew Petach wrote:
> blocks.  If I have 64 non-contiguous /24's to return, and am asking
> $1,000 each for them, but all ARIN has in the bid pipeline is 8 bids
> for /22's at $4,000 each, should ARIN
> a) spend $64,000, purchase the /24's and hold them in case future
>     /24 requests come in, but reject the bids due to lack of /22s
>     (leaves ARIN $64,000 poorer)
> b) spend no money; reject the offer of space, and reject the bids
>     due to lack of /22s
>     (no net loss or gain in the fund)
> c) accept the bids as an open-ended offer, and keep them on the
>     books, potentially for months, until /22 blocks become available,
>     adding them either 1) sequentially into the wait queue for /22's, or
>     2) adding them by bid amount into the wait queue for /22s

As written right now the policy requires c, but specifies the bids
are good for 60 days at a time unless renewed by the bidder.

Regarding C1 and C2, well, it's complicated.  We like a first-come
first-served model, that's the way it has always been, and it seems
fair.  However, let's look at the following bids, in timesequence order,
for a /22:

   $500, $600, $400, $800, $250, $900.

ARIN finds two folks with /22's willing to return them, but only at a
price of $750.  It seems like the correct answer is to give them to the
$800 and $1000 bidders, and then re-sort the queue to be:

   $500, $600, $400, $250

The policy attempts to specify this, however having thought about this
for a really long time, when you add in deaggregating and aggregating
space the algorythm becomes really, really complicated.  If you go back
in the PPML archives you'll find a bunch of folks proposing algorythms
in the original discussion on this.  None of them are fool-proof.

In this draft I've laid out general guidelines and left the rest
to staff as an operational matter.  This isn't taking the easy way
out, on the contrary I think the rules will have to have minor
adjustments over time as the policy is actually implemented.  The
weird corner cases will come up, and a call will have to be made.

Now, back to your example.  You've picked a particularly hard case.
The only way ARIN can allocate a /24 is via the micro-allocation
policy, and while I don't know the exact rate of applications I
suspect it is extremely low.

However, let me point out something this policy allows.  Let's say
you hold 10.0.0.0/24, 10.0.1.0/24, and 10.0.2.0/24 and are willing
to sell all three.  I hold 10.0.3.0/24, and I am actively using it
and unwilling to sell the space.

ARIN has the ability, via the current contract to take back numbers
and issue new ones.  I don't know if they have ever done it or not,
but they can.  I'm not suggesting they just do that, as it would
be rude and likely not what the community wants.  But let's try
this...

ARIN has a bid for $6k for a /22.  ARIN offers you the $1000 asking
price for the three blocks you hold, and offers me $500 to return
my block at which time ARIN will replace it with a different block
(say 10.0.4.0/24, which happens to have never been allocated for
some reason).  In essense they replace my block, but pay me for the
trouble of renumbering.

ARIN now has 10.0.0.0/22, at a cost of $3500.  They add say $500
for the staff time to figure this out, make the phone calls and all
that and sell it to the bidder for $4k.

In essense the money can also be used to pay someone to move to
"defragment" the address space.  I suspec the payment to move will
be significantly less than the payment to just return space; but
none the less it recognizes that renumbering into a new block so
someone can have that /22 costs time and money, and compensates for
it.

For a /1 to a /22 I suspect ARIN will have orders of magnitude more
bidders than they do address space, at least after a short period
of time.  As the address space gets more efficiently used there
will simply be less and less folks willing to part with it.  Thus for
those sizes I would think ARIN would always have a deep pile of bids.
The issue comes in with /23-/24, which today can only be given out via
micro-allocation policy.  I think the community needs to consider this
sub-issue carefully no matter what policy is adopted, there are plenty
of "swamp space" /24's out there that may appear in the market in one
form or another.

> will be happy to point out.  It also puts us right back into the "big
> guys with money get all the space" scenario that people are so
> fond of whining about here.  (disclaimer--I work for a big company
> with money.)

In all of the schemes I have seen proposed the guy with the money
gets the space.  Frankly, I don't see a away around that.  However,
some of the proposals attempt to make an interesting restriction
on the big guy, including this one.

If you come to ARIN and get qualified for a /12, then in this policy
you get to bid on a /12, period.  ARIN can't fill your order with
a /11, /10, and two /9's.  You must wait for a /12 to come up, or
something larger ARIN doesn't have bids for so they are willing to
subdivide.  One of the worries is someone comes along and justifies
a /8 (say, a cell provider) and hoovers up all of the little chucks
of space all over.

-- 
       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/20090411/27e607ec/attachment.sig>