Re: [sig-policy] IPv4 countdown policy proposal
At 03:43 07/02/13, David Conrad wrote:
>First, I'd like to thank the folks at APNIC for initiating this
>policy discussion. I should also note explicitly that this message
>is sent in my personal capacity, NOT as a statement from IANA or
>ICANN. These are only questions/comments from me, nothing more.
>
>With respect to the proposed policy, I have the following comments:
Thank you for your useful comments.
First of all, I believe the discussion should be done
based on comparison between this policy or "doing nothing special"
or some other counter proposal.
Everyone can easily point out the proposed policy would give pains to ISPs
and it is actually very sure. But no such "count down policy" would also
give pains, as we believe, much more pains. So, one of goals is
how to minimize such pains. We do beilive the proposed policy can minimize
them to ISPs by letting them know the T-date in advance.
>> (1) Global synchronization:
>> All five RIRs will proceed at the same time for measures on IPv4
>> address exhaustion.
>
>I assume this synchronization is in the context of allocation policy.
>
>Given the timeframes we're looking at, it isn't clear to me the
>current RIR policy definition mechanisms are agile and/or quick
>enough to adjust to the changes in the address "market" in a
>synchronized fashion. Would it make sense for the RIRs to come up
>with an expedited policy definition mechanism?
>
>As each RIR is an independent actor required by their respective
>bylaws to be responsive to their communities, how would such global
>synchronization be enforced? Also, wouldn't such synchronization in
>the face of increased demand due to scarcity be considered collusion
>or cartel-like behavior?
Good point. I am understanding that global coordination is very dificult.
But I believe this is not a kind of issue "we will give up global coordination
becuase it's hard" but "we should really need global synchronization and
we will try to seek how to achieve it."
I expect NRO and ICANN ASO to play important roles to it.
>> (2) Some Blocks to be left:
>> Keep a few /8 stocks instead of distributing all.
>
>Obviously, this would result in the IPv4 free pool runout date being
>artificially accelerated. I'm not sure the justification as provided
>in the policy proposal supports this acceleration. Specifically:
According our policy and itsestimation, less than ten /8s (maybe 5, 6 or something)
would be left. This is amount of less than one year usage (current usage
is 10-13 /8s per a year globally)
As you pointed out, this policy would artificially accelerate the runout date,
but it's just 6 months or so. Is this a significant difference?
>> It is expected to cause
>>confusions
>> if one party can receive an allocation while the other has to
>> give up, just with a touch of a difference.
>
>This is simply reality and (I'm told) has been experienced by folks
>already. I would imagine more confusion (and heated discussion)
>would be generated by the existence of a remaining "reserved" free
>pool when people are turned away who could have been satisfied with a
>portion of that free pool.
>
>Additionally:
>
>> requirements to start
>>a translator
>> service between IPv4 and IPv6 networks should be supported, and
>> there may be some requirements in the future that are beyond
>> our imagination at this current moment.
>
>Realistically, any sort of protocol development of this nature would,
>by necessity, have to deal with the fact that large blocks of
>contiguous IPv4 address space are unavailable. Proposing to reserve
>blocks of addresses for some prospective and unimagined future
>development raises the question of "how much to reserve?" and I don't
>see a particularly good answer to this question.
I don't have a good answer, either. Probably no one knows.
But this cannot be a reason we don't need some blocks to be left.
>> (3) Keeping current practices until the last moment :
>> Maintain the current policy until the last allocation.
>
>It is unclear to me that such an approach can withstand the demands
>of the address-consuming community. Fundamentally, the issue the
>community is facing is the same as faced by any consumer of a limited
>resource and I believe there are textbooks written on Scarcity
>Theory. IANAE (I am not an economist), but I believe the two ends of
>the spectrum in dealing with scarce resources are:
>
>a) centralized control with rationing ("to each according to need")
>b) marketization
>
>What this policy proposal is suggesting is a rather extreme form of
>(a), with little regard to what happens after the last IPv4 block
>from the free pool (either the hard limit or the artificial soft
>limit created by reserving some /8s as described in principle 2) is
>allocated. I am skeptical the demand for IPv4 addresses is going to
>magically vanish on that day, rather I suspect ISPs will do what is
>necessary to ensure they are able to meet their customer's demands
>(see (b)).
>
>If the community wishes to continue with centralized control,
>wouldn't a graduated approach to rationing (e.g., increasing the
>amount of assignments needed to justify an allocation over time) be
>preferable to a binary "yes we have addresses/no we don't" flag day?
Yes, this is worh discussing.
If someone can propose a graduated approach, then we can compare it
with the proposed policy.
But in considering the policy, we simulated this kind of alternative.
For example, assume a graduated approach can save 20% of current usage.
It would result in only one year's life extention. On the other hands,
20% conservation seems very tough to ISPs. ISPs would have to pace down
their business without transitioning to IPv6 anyway.
Regards,
Takashi Arano