[ppml] Policy Proposal: 2007-12 IPv4 Countdown Policy Proposal

[Stephen Sprunk]

Thus spake "Leo Bicknell" <bicknell at ufp.org>
> To that end, I'd like to oversimplify the proposal.  Language,
> format, and justification aside I believe the proposal can be
> boiled down to the following simpler statement:
>
>     The RIR's, in order to assure the orderly shutdown of IPv4
>     allocations should do their best to predict the date at which
>     there will be no more IPv4 addresses available, should
>     announce a termination date just before the predicted
>     exhaustion, and should cease allocations on that date even
>     if there is some address space still available.
>
> I believe the intent of the authors is to realize a number of
> potential benefits:
>
> - There is a well known date at which no more IPv4 space will
>   be available, making it easier for those needed addresses to
>   show their management the need for alternate plans.

Also, having the RIRs jointly announce such a date gives it credibility.  A 
few slides by Geoff or Tony, as much as we respect their work, doesn't have 
the same impact as an official announcement.  That is something that can be 
presented to management, vendors, customers, etc. to try to motivate them.

Personally, I don't think it'll motivate most of them enough to be ready in 
time, but at least they couldn't blame us for not telling them.

> - By the RIR's shutting down distributions of addresses at the
>   same time it prevents the "last RIR standing" from being
>   swamped by every international company solely because
>   they still have addresses.

I think this is handled well enough under existing policy.  An org with an 
int'l network is _supposed_ to go to the local RIR for each part of that 
network, and the RIR boundaries are such that enforcing that rule is not 
burdensome.  OTOH, I'm not naive enough to think that people don't go 
RIR-shopping for favorable policies, just like lawyers go judge-shopping 
before filing a case.  However, the vast majority of orgs only deal with one 
RIR; an ISP only in the US is going to have a really tough time convincing 
RIPE or LACNIC to allocate it a /16 after ARIN runs out.  For truly int'l 
orgs, well, what's the harm?

> Of course, there are drawbacks:
>
> - This requires global coordination.
>
> - We may leave some IPv4 space unused that could otherwise
>   be put to good use.

That's why I'd prefer that the IETF and/or IANA mark space (a /8 or two) 
explicitly reserved for uses such as 4to6 gateways and other future IPv4 
protocols (and let's face it, there will be) that require well-known 
addresses, and the RIRs should allocate whatever is left, down to the last 
address.  The announced termination date will have to be given as a best 
estimate anyways, since projections will not be able to take into account 
the inevitable change in behavior after the date is announced.

This has the (unfortunate?) side effect that even if IANA stops giving out 
/8s to RIRs on a given date, some will likely run out _months_ before 
others.  LACNIC and AfriNIC don't go through a /8 anywhere near as fast as 
ARIN does -- and I don't believe we should tease folks in those regions by 
giving them an /8 and then telling them they can't use most of it just 
because folks on other continents have a problem.

> - This policy itself may cause a run on IP space.

Indeed it will.  People are going to lie, cheat, and steal their way to 
addresses in the months before the big cut-off.  Other folks who currently 
only apply for new allocations once or twice a year will start doing so 
monthly.  Folks with many existing allocations will suddenly take advantage 
of the trade-in policies en masse to get additional space via rounding. 
etc.

One must always remember the law of unintended consequences when trying to 
regulate people into something they won't like and probably won't feel is 
just.

> There are alternatives, Owen DeLong just wrote about what
> would probably be considered the opposite viewpoint in
> another message, I quote:
>
>     I believe that the system will function and that there is no need
>     to  do anything different until ARIN is unable to fulfill requests.
>     At that time,  ARIN should fulfill request it can on a
>     first-come-first-serve basis and provide  a polite apology in
>     response to requests which cannot be fulfilled.  I do not  believe
>     a change of policy is required in order for ARIN staff to do this.

For the record, I support Owen's position.  Not because the end game is 
pretty, because it obviously isn't, but because it makes no sense to me 
that, as addresses are returned to (or reclaimed by) ARIN, we wouldn't be 
able to hand them out again.  I also don't like the idea that folks wouldn't 
be able to use the trade-in policies, if ARIN happened to have blocks large 
enough to, for instance, consolidate the announcements of someone who 
currently has several hundred discontiguous blocks.  Freezing the IPv4 
registry on a particular date isn't necessary.  ARIN (and the other RIRs) 
can still do many useful things even if there's no more "new" space left. 
And we, as policymakers, can change policy to free up more "used" space for 
new allocations/assignments if we decide to do so.

> Last, in an attempt to keep the discussion focused, I'd ask you
> to consider if these related topics are relevant to this policy's
> thread, along with why I think most are not:
> [snip]

I agree with you that all of the issues you list are relevant to ARIN but 
not to this particular proposal.  I would definitely like to see distinct 
proposals for each of them, and some lively debate around those proposals, 
which can be done concurrently with this debate.

S

Stephen Sprunk      "Those people who think they know everything
CCIE #3723         are a great annoyance to those of us who do."
K5SSS                                             --Isaac Asimov