[arin-ppml] quantitative study of IPv4 address market

Owen DeLong owen at delong.com
Tue Sep 4 18:03:17 EDT 2012

On Sep 4, 2012, at 13:18 , Joe Maimon <jmaimon at chl.com> wrote:

> Owen DeLong wrote:
>> On Sep 3, 2012, at 08:24 , Joe Maimon <jmaimon at chl.com <mailto:jmaimon at chl.com>> wrote:
>>> Owen DeLong wrote:
>>>> On Sep 1, 2012, at 20:44 , Joe Maimon <jmaimon at chl.com <mailto:jmaimon at chl.com>> wrote:
>>>>> Owen DeLong wrote:
>>>>>> The goal of ARIN address policy is to place addresses in service where they are needed so long as that is possible.
>>>>>> This is the overarching goal of both the allocation policy _AND_ the transfer policy.
>>> The allocation policy is relevant only so long as ARIN has an allocation pool. Which I want to see last as long as possible, since it is certain not to last long enough.
>> This is where we utterly and completely disagree. Making the free pool last artificially longer by disadvantaging legitimate uses of the address space today is not a win and is contrary to ARIN's mission statement, IMHO.'
> Legitimate uses and justified need are equivalent statements that describe demand the community can support with supply, which is affected by scarcity, and that has always been the case. As such, the ARIN free pool is still wholly available to legitimate uses of resources, just like always.
> The bar has changed, just like it has in the past.

The 3 month policy neither raises or lowers the bar. It merely creates a limitation on one particular class of users.

Alone, i would consider it not terrible, but not particularly desirable, either. I will note that I was among those that didn't think it was such a bad idea when originally considered by the community. We have more information now.

However, when combined with a 24 month window on transfers, it is particularly unfair and counter-productive to ARIN's mission.

>> I appreciate the benefits of the transfer market but also recognize its limitations and disadvantages. Inflicting those limitations and disadvantages on some class of users earlier in order to delay it for other classes of users is not good policy and as it does not meet the required fairness test.
> Yes it is and yes it does. Or do you think the transaction we have seen should have been done from the free pool? And if not, does that mean that the transfer policy should be the focus of your ire, instead of the free pool?

There is no particular ire here in any direction, so your statement above does not make sense.

How is it fair to force one subclass of users to the transfer market in order to preserve free pool for another subclass of users? Please explain to me the logic by which you determine that this is fair to both the users denied access to the free pool.

The mere assertion that it is is not sufficient here.

I don't know whether those transactions should have been done from the free pool.

>>> Apparently to the extent measurable by the ARIN allocation fall-off not made up for in transfer market.
>> More than likely this is a temporary rather than permanent phenomenon. Most likely many organizations are postponing additional requests due to the increased difficulty of obtaining those addresses which will eventually lead to a greater crush of supply at a later date.
> If it is still occurring, it would be a year long temporary phenomenon.

Your point being? There are signs that free pool consumption is beginning to ramp back up, by the way.

> Postponing requests due to increased difficulty results in negative benefit and is irrational behavior, unless you mean increasing efficiencies to compensate for the additional difficulties, in which case, I say again.

You say efficiencies, I say network dysfunction.

> Working as designed.

Not at all. Design included end-to-end addressing.

>> Does it? I don't know how much supply remains at achievable prices and I'm not sure what you are looking at that tells you it is vast or that more than $15 per address is necessarily an achievable price point.
> If you assume that even half of the un-routed space is available for the market, or that half the legacy is available, or that a third of the broadband consumer addresses are re-usable or available for the market, you begin to doubt very much that the address market has any short-term hard limits.

I do not assume that half of the un-routed space is available for the market. Why would I make such an absurd assumption. The vast majority of un-routed space is used in networks which connect to networks which connect to the internet but which are not connected themselves.

Why would I assume that half of the legacy space is available? I've seen nothing at all to support such a claim.

Why would I assume that 1/3 of the broadband consumer addresses are re-usable or available for the market? I've seen nothing to indicate any truth to such a claim.

> We do not know how much inefficiency is available to be wrung out, but I have no reason to believe it be anything less than 25%.

End-to-end addressing is _NOT_ inefficiency.

> Under a reduced by scarcity and market pressures utilization rate, that can last years.

Let us hope not. Those will be very painful years.

>> No, my expectation is that eventually, IPv6 must replace IPv4 as the lingua franca of the global internet because there is currently no other available solution which will allow the internet to continue to grow and provide the same vital functions that it provides today. My hope is that we will do that sooner rather than later because it will be much less traumatic, much less disruptive, and much less expensive.
> We are so far beyond that point that it is salt in the wound to hope for it.

I completely disagree. The current curves on IPv6 adoption imply that without additional pressure, IPv6 will be close to ubiquitous in 5-10 years. IPv4 runout will provide additional pressures and probably accelerate that timeline somewhat.

>>> Do you have reason to be hopeful or is that simply what you have banked on and wish to happen?
>> Yes, I have many reasons to be hopeful...
>> 1.IPv6 adoption is growing
>> 2.Growth in IPv6 adoption is accelerating
> Project from this growth, rate when the demand for IPv4 addresses will die off. Put a number on it. Then let us agree to try to conserve the ARIN free pool to last that long.

That's nonsensical.

What is the point of having a free pool when there is no demand for addresses? By definition, all addresses remaining in the free pool at that time are wasted.

Addresses retained in the free pool are UNUTILIZED addresses. This is the ultimate in inefficient utilization.

The goal of ARIN policy is to make resources available to legitimate uses to the greatest extent possible. This is not achieved by hoarding addresses in the free pool while demand for them exists.

>> 3.No matter what you do to try and work around it, IPv4 cannot scale to support the internet much beyond its
>> current size. Efforts to do otherwise have been degrading the capabilities of the internet for more than a
>> decade now
> Nobody has been switching to IPv6 because of this, so what makes you think they will soon?

What do you mean nobody has been switching to IPv6? Many people have been deploying IPv6 and more are continuing to do so every day.

>> and all evidence is that this will continue to get worse and not better going forward. Many
>> indications are that it will get rapidly and substantially worse.
> This is your best case scenario? That IPv4 will become so untenable that it will force IPv6 adoption?

No, that is my worst case scenario. The best case scenario is that we move towards IPv6 before IPv4 reaches that point.

> Unfortunately for this scenario, there is still no surety that there will exist a first mover advantage, even then.

There already exist first mover advantages. Hurricane Electric is reaping many of those first mover advantages.

However, we are so far past the first-mover stage of IPv6 deployment that looking for first-mover advantages doesn't make much sense.

>> 4.There is strong evidence to suggest that IPv4 is unsustainable and that any of a variety of run-out related
>> factors and activities will eventually make IPv4 utterly unusable.
> Eventually the sun will burn out and Earth will be utterly uninhabitable. I suspect sometime before that, we will have migrated completely off of IPv4.

The difference here is that the time lines are radically different. IPv4 will become unusable almost certainly in less than 10 years and very likely in less than 5. OTOH, the sun has enough years remaining that the entire IPv4 address space is not enough to assign one to each year. IPv6, OTOH, would last significantly longer. ;-)

> But I dont know when that will be. Neither do you. And as such, we must not treat IPv4 as disposable.

We don't know when it will be, but it is pretty clear that it will be relatively soon. I am not advocating treating IPv4 as disposable. Preserving a free pool while starving valid demand for addresses simply doesn't make sense.

>> It equates to all of the RIRs being out at similar times. This is actually advantageous as it is less likely to push for asymmetrical deployments of IPv4 continuation technologies creating a more drawn out transition process with greater pain and dysfunction.
> Horse trade? You support APNIC/RIPE style last /8 conservation policies and I will support 12 month free pool allocation.

While I don't think such a policy is the best policy, that combination would certainly be better than our current policy.

Back in the day, I actually did support something similar. As a result of community pressures, it was reduced from /8 to /10.
Feel free to review the history. It's NRPM section 4.10.

You will notice that it was originally suggested as the full /8.


>> I would call that the ultra-cynical view and also one which ignores the following realities:
>> 1.Most growth in the ARIN region is in the mobile market.
> Which is the most CGN targetable market. The cynical view is that the growth serves as a sanctioned form of stockpiling.

An odd suggestion at best.

>> 2.VZW and T-Mobile have put huge effort and investment in moving their mobile
>> solutions towards IPv6.
>> 3.Comcast, Verizon, and AT&T have all put significant investment in making IPv6
>> available to their broadband subscriber base and are continuing to expand that
>> offering to more of their subscribers.
>> 4.Many other broadband providers are engaged in similar efforts.
> Do you think the ARIN free pool continuing to be available to their competitors may  have something to do with it?

I tend to doubt it.

> And they can safely plod the course. It will still give them years of advantage. From availability, to supported, to desired, to consumed. Quite a distance.

Not so long as you seem to think, IMHO.

>> 5.The majority of large legacy holders are not ISPs.
>> 6.The majority of ISPs that are large legacy holders hold more non-legacy space than
>> legacy space at this point.
>> 7.Nobody who has looked at the technology seriously wants to run a large-scale
>> CGN deployment in a country subject to CALEA or CALEA-like legislation.
>> 8.Nobody who has looked at the technology seriously from an operations perspective
>> wants to maintain a CGN deployment at any scale.
> You are correct. From an ops viewpoint, it sucks. Does it matter? Who calls the strategic shots? Ops?

From any of the following viewpoints:
	1.	Operations
	2.	Support
	3.	User Experience
	4.	Marketing
	5.	Sales
	6.	Legal
	7.	Cost
	8.	Accounting

It pretty much sucks. Surely someone somewhere in that list has at least some significant input into the decisions in each organization.

>>> We will not be better off. There is no way to predict how long IPv4 will remain relevant and/or necessary.
>> We will be better off in that we will not extend the duration of IPv4 relevance to cause greater pain and
>> disruption.
> IPv4 relevance is not decreased by the exhaustion of RIR free pool. RIR relevance is.

We can agree to disagree about this.

>>>> Extending the duration of the ARIN free pool by creating an early artificial shortage through policy
>>> Extending the availability of the ARIN free pool by reducing its burn rate is the only sane approach.
>> Extending the availability of the ARIN free pool is impossible.
> It is happening right now. So it is possible.

No, it is not. The availability has been reduced in one direction while extended in another. The total availability, however, continues to shrink.

>> What you are talking about is extending
>> the duration of the ARIN free pool by reducing its availability and making it asymmetrically and unfairly
>> unavailable to various classes of users to the benefit of a class of users you happen to advocate.
> Lets not discuss fair. Because then we have to talk about things like who has the addresses and who does not. And who pays more and who pays less. And on whom the burden of obtaining these addresses is greater. And on whom is policy more burdensome. And who tends to find working with ARIN to be more difficult, frustrating and mystifying. And whom reaped the benefits of rir history that relative to now is increasingly more relaxed and informal.

I understand that you would like to avoid fairness as a topic because it does not support your position. However, it is a primary factor that I am obliged to consider when considering possible ARIN policies.

> Instead, lets discuss legitimate use. Justified utilization. Documented need. All things which ARIN, both operationally and as a matter of policy has gradually and ever increasingly been raising the bar on. For years. They certainly believe that they are applying policy soundly and fairly.

They are we. If you raise the bar for everyone, I'm fine with it. If we went to a universal 3-month supply policy across the board (transfers, ISPs, end-users), I would have no objection. If we went to a universal 12-month supply policy across the board (transfers, ISPs, end-users), I would have no objection. Personally, I think this is probably the best compromise. I'd even be willing to accept a universal 24 month policy.

It is the dichotomy of duration in these different policies that is inherently unfair. That dichotomy is absolutely a new phenomenon that has not been present for years.

> The result is that ARIN still has a free pool to discuss.

That is one of many results. Certainly it is the only one of the list of results one would want to point out if one were attempting to support current policy.

>> As a community member, I can somewhat sympathize with your view even though I find it misguided.
> Likewise,  I can sympathize with your desire to decrease IPv4 relevancy. However, I find your attempts to do so misguided and coy.

I am not attempting to decrease IPv4 relevancy. I am, however, attempting to decrease the duration of the transition period because the level of pain associated with transition will only increase over time. The longer transition takes from this point, the more painful it will become.

I don't believe I have been coy in any way. I have attempted to make my reasons and my position quite clear. I think you are the first person in the ARIN community to accuse me of shyness or reluctance to express details of my position.

As to misguided... I find your attempts to extend the duration of the free pool by further and asymmetrically reducing its availability to be misguided.

>> While the above thoughts are my own personal opinion and do not represent an official statement of
>> the AC (or even agree with the official opinion of the AC), my personal thoughts on this matter are
>> partially shaped by my obligation as an AC member to insist that policy be fair, technically sound,
>> and useful to the community.
>> Owen
> The AC has no business having an official opinion in this matter (and quite a few others) but that does not seem to stop them, so I would not worry about that if I were you.

In any matter of policy, the AC absolutely has business having official opinions. However, it is not my role to speak for them unless asked to do so by the chair and I don't speak for them.

My point was that my personal opinion is, in part, shaped by the fact that as an AC member, I have to seek fairness in policies.

The current policy is, IMHO, unfair and should be corrected.


More information about the ARIN-PPML mailing list