[arin-ppml] ARIN-prop-162 Redefining request window in 4.2.4.4

Owen DeLong owen at delong.com
Sun Jan 29 12:21:54 EST 2012


On Jan 29, 2012, at 7:45 AM, Joe Maimon wrote:

> 
> 
> Owen DeLong wrote:
>> On Jan 28, 2012, at 7:53 PM, Joe Maimon wrote:
>> 
>>> Based on John's numbers, a /12 a year is more than enough. I believe 8-10 years post IANA runout of guaranteed resources for new entrants for the cost of /9 is a quite reasonable and respectable behavior for a public resource stewardship entity to be engaged in.
>>> 
>> I suppose that's one way to look at it. Another way to look at it is that preserving a /9 for those that may exist vs. using it for those that DO exist is profligate waste of resources.
> 
> I dont think that word means what you think it means.
> 

I'm pretty sure it does.

> Balanacing the history of consumption against 0.2% of the consumed resource to be used in a far more responsible fashion, I fail to see any rationale to term the latter profligacy.
> 

I think, however, we disagree on the contextual definition of the term responsible.

> They didnt gorge themselves enough yet, is that it?
> 

ISPs do not consume addresses primarily for their own purposes. They use them to provide services to their customers (mostly by handing those addresses out to said customers for use on their equipment).

Denying addressees to customers that want to use them today through existing providers in order to keep them available for possible customers later who might use them through providers who don't exist yet is not, IMHO, responsible.

>>> 
>>> This I dont get. What issues? We should all be equally miserable? Is this a race to the bottom?
>>> 
>> The issue is the various non-solutions that it spurs in the areas where runout has occurred early which then create additional costs to transition.
> 
> So now your are a proponent of helping those who have no resources at the expense of those who do? How does increasing our consumption rate of a still available resource do that, unless you are proposing ARIN transfer its free pool to APNIC?
> 

Interesting interpretation of my words.

Let me try and clarify my thinking in the hopes of actually communicating with you rather than continuing to engage in heated rhetoric...

We have entered a period of escalating pain. The pain will continue to escalate until the transition to IPv6 is complete. The sooner we complete that transition, the sooner the pain will end. The longer we hold out and attempt to cobble various solutions to keep IPv4 limping along, the more pain we will inflict on ourselves and on the global internet. The period of pain began almost 15 years ago with a relatively minor scratch (NAT). We've allowed that scratch to fester, expand, and become pervasive where there are now more end-users behind NAT than on the actual internet.

Now, we're talking about pushing these pervasive minor lacerations into deep flesh wounds in the forms of CGN, NAT64, and other solutions that provide a progressively less functional internet at progressively greater cost.

> It will runout. There is no stopping that. Those for whom it ran out faster than those who it didnt have done a poor job in stewardship and in preparation, and I dont see why we have to express contrition and repentance towards them, at our own expense.

APNIC has done a poor job of stewardship? Seriously? You're talking about more than 50% of the world population in the APNIC service region, yet they have less than 25% of the IPv4 address space. How can you possibly claim that is a poor job of stewardship?

OTOH, ARIN administers more than 50% of all IPv4 address space and yet our region encompasses less than 25% of the world population and yet only the smallest registries with the poorest regions have any chance of running out later than we will even under proposal 162.

Stewardship means getting the addresses that are available into the most effective use possible. It does not mean preserving a free pool for unknown possible uses in the face of known scarcity for legitimate present uses.

> Can you stop hand waving and actually be specific? Are you referring to APNIC? Are you referring to US mega ISP's who are finding it slightly more difficult to scoop up ipv4 one /12 at a time?
> 

I'm not referring to any of them specifically, but, the overall global situation with the transition from IPv4 to IPv6 and ARIN's role in that transition.

> 
>> 
>>> The only way to get even runout is for all the RiR's to decide upon a date after which aint nobody getting nothing.
>>> 
>> No, there are actually several other possible ways to even things out. We're not talking about perfectly even, but, several years of asymmetry is a bad thing.
> 
> How so? Your cure is worse than the alleged disease.
> 

We can agree to disagree about that.

> Let APNIC go on an aggressive reclamation and reservation policy regime until their runout matches ARIN's again.
> 

I would be willing to bet that there is far less underutilized space in the APNIC region than in the ARIN region.

> Since it is their problem, then it should be their impetus to resolve it.

While I would agree that they are presently the region suffering the most, your claim that it is "their problem" ignores the genesis of the problem, the history of the internet, and the role of ARIN and its predecessors in the creation of that problem. See the statistics above.

> 
>>> Precisely what does that solve?
>>> 
>> See Geoff Huston's presentation from Philadelphia and/or Busan.
> 
> Geoff's presentation shows that ARIN's policies are working to preserve IPv4 access, APNIC's policies have failed, IPv6 transition has failed and shortly after stating he did not know the solution, he falls into the trap that increasing and hastening our collective misery will somehow drive IPv6 and not make everybody upset at us for our poor judgement.
> 

An interesting way to characterize what he said. Another way to look at it is to realize that IPv4 global runout is upon us. It is inevitable in all regions. The longer it takes, the longer it will take us to complete the transition. Said transition has not failed, it is merely not proceeding as rapidly as would be ideal. Continuing to limp along with the present state of IPv4 is painful. Attempting to preserve it longer will become progressively more painful and more expensive. Focusing on the transition instead of limping along with IPv4 will create a small increase in short-term pain levels, but, provide much greater long term relief much sooner.

> I disagree on both counts.
> 

As is your right. Time may tell which of is is more correct, or, time may only tell us that whatever path we do choose as a community was obviously not ideal.

> Why dont we all just turn on the IPv4 evil bit on an agreed upon date then?

Had we done that 10 years ago, that might have actually been a better idea. It certainly did a better job of facilitating the transition from NCP to TCP/IP than what we have done with the IPv4->IPv6 transition.

> 
> What Geoff fails to discuss in the presentation I saw is that for the End User at the edge (which he does acknowledge as the transition challenge), runout is not defined at IANA or RIR levels. It is defined by the SP. And the SP will always have IPv4 resources, to divvy up in a manner most efficient and value driven as per its own judgement.

Quite frankly, the end-user transition is going to occur fairly soon whether we want it to or not. The only way SP can continue to have IPv4 resources to divvy up is by providing a progressively more and more degraded user experience to their customers, so, your claim there is utterly inaccurate.

OTOH, if the end user transition is overtaken by events and occurs before there is sufficient critical mass among the service and content providers on IPv6, it will be much more painful for both the end users and their service providers.

While Geoff may not have covered that fact in detail in the slide deck, I believe he at list mentioned it in his presentation.

> So long as they arrived in time for the party while the getting was good.

Even if you got to the party early and got a /8, once you pass the 16,000,000 customer mark, things are likely getting pretty tight for you even if you only give one address to each customer (which is already a degraded form of service). Beyond that, all you can do unless you can get more address space is to further degrade your services.

> Exhausting RIR stock serves only to harden and restrict the pool of the available service providers, to deny qualifying end users any other option than being at the mercy of the collective service providers with a monopoly on the resource, to the mercy of the market when they are the smallest, most vulnerable participants in it.

On the other hand, preserving a free pool and preventing those existing providers from expanding their service offerings without degrading the quality of their services will deny qualifying end users any other option than progressively more degraded internet experiences and force those service providers to inflict those degraded service levels not only on new customers, but, also on some (or all) of their existing customers as well.

Bottom line, no matter how you manage it, runout hurts the end users.

> It certainly is not stewardship of a public resource, even as it may be the appropriate end state of private property.

Stewardship of a public resource is a balancing act to try and get the resource(s) into the most effective use. The free pool cannot possibly be the most effective use of a resource in a time of shortage.

> Fail.
> 
> (He does confirm what all the non-optimist have been saying for years, those whom you have claimed to not number yourself a member of all that time. Transition has not occurred with IANA runout. It has not occurred with RiR runout. It will not occur with SP runout, nor has it been induced by CGN and the like, which was our last hope. It will only occur with a seamless and properly working backwards compatible transition technology that doesnt eat its own expense via its own success. Fix 6to4 or Teredo or replace it, otherwise I can only see modest pickups in transition speed for the next decade.)

Transition IS occurring. It has been occurring and it will continue to occur.

Transition speed has actually been achieving a more than modest pickup in speed since last February, so, your statement here is not born out by the statistics. The problem is that even with that acceleration, it's still too little too late for avoiding significant pain in the transition process.

There is a backwards compatible transition technology that doesn't eat its own expense via its own success. It's called dual stack. It's the only 100% viable transition technology that does not damage the user experience. Unfortunately, it requires IPv4 availability which we cannot preserve at this point.

> 
>> The presence of a transfer market alongside a free pool causes issues. It is the interaction and combination of the two factors that is creating concern. Yes, transfers have their own set of new problems they bring to the table, but, that's an unfortunate necessity of the current state of things.
> 
> Stop handwaving and start with the specifics. I dont see any issues caused by the existence of both that are not made worse by the elimination of the other.
> 

1.	There are tremendous potentials for abuse of either by leveraging the other.
2.	So far, the market has only served to prevent the return of addresses to the
	free pool from bankruptcies, thus increasing the cost of IPv4 resources while
	not actually increasing their availability.

There are other specifics that I cannot go into due to NDAs.

>>> I choose slower ARIN resource utilization by those that got while the getting was good, enabling those who did not to still obtain them without subjecting them to the potential intractability of the address market, fueled by those who did.
>>> 
>>> Thats good stewardship.
>>> 
>> What you call good stewardship, I call a form of socialism.
> 
> Its called many things, most of them good. Good business sense to ensure a steady influx of new customers. Good stewardship of a public resource that has been consumed in profligate fashion in years past. Good survival instincts to remain relevant to the needs of your constituency and to publicly show responsibility and even handedness, even if belated.
> 

But this doesn't ensure a steady influx of new customers. Instead, it ensures that we force existing customers to choose between not accepting new customers or degrading the services to their existing customers in order to support new ones.

There is little or nothing that can be done about past profligacy. However, whether a 3 month or 12 month needs basis is used in ARIN policy, I do not believe you can call the present needs-based policies profligate in either case. At least not with any degree of accuracy.

Frankly, experience with the 3-month policy to date has shown the opposite of remaining relevant to the needs of our constituency. The policy is equally even handed whether it is based on 3 month or 12 month need, so, I fail to see your argument there.

> Not good for those who still want gorge as they have in the past. Fetch me a violin, please.

Sorry, not buying this.

> Now, community networks for free or reduced fee, that sounds more like it. Correct me if I am wrong, but didnt we both support that?

I still support that. You will, however, note that the community networks policy (which did not get the fee support we hoped from the board),
applies only to IPv6 and not to IPv4.

Owen

> 
>> 
>>> IPv6 relevancy to the consumption of IPv4 has been vastly overstated to date.
>>> 
>> Huh?
>> 
>> Owen
>> 
> 
> Please follow along with me for a bit.
> 
> The existence of IPv6 has not made IPv4 any less relevant or in demand. The runout and exhaustion of IPv4 has not may IPv6 any more then marginally more relevant and in demand.
> 
> Contrary to those who confidently claimed it would be otherwise.
> 
> Was that not your whole thrust, specifically the inclusion and quoting of Geoff's research to bolster your assertion that all of us exhausting faster is good for the internet, by which you mean the IPv6 version of the internet?
> 
> It most certainly is not good for the IPv4 internet, which is the one we have now.
> 
> To reiterate. Slash and burn of IPv4 in the pursuit of IPv6
> 
> a) unwise
> 
> b) unsound
> 
> c) unkind
> 
> d) unhelpful
> 
> e) unsupported
> 
> f) all of the above
> 
> Best,
> 
> Joe




More information about the ARIN-PPML mailing list