[arin-ppml] Fw: A Redefinition of IPv4 Need post ARINrun-out(was:Re:Against2013-4)
Mike Burns
mike at nationwideinc.com
Thu Jun 13 17:01:56 EDT 2013
Hi Jason,
My reply to you was off-list, but since some others have replied on-list, I will also send this one to the list for consideration.
Regards,
Mike
From: Mike Burns
Sent: Thursday, June 13, 2013 3:04 PM
To: Jason Schiller
Subject: Re: [arin-ppml] A Redefinition of IPv4 Need post ARINrun-out(was:Re:Against2013-4)
Hi Jason,
Do you really think that small ISPs seeking to fund 40 years of IPv4 address needs pose a problem that is greater than the problems posed by retaining the needs test for all transfers? Remember we are talking in the context of a /12 annual cap or a /12 aggregate cap on needs free transfers. I fail to see the threat posed by Mom and Pop ISPs with $10 million on hand to ensure they will have IPv4 addresses in 40 years. I doubt you could identify a single likely candidate for this group.
There are real, extant problems associated with retaining the needs test for transfers, chief among them driving transactions off the books to the detriment of the one principle everybody has agreed on- the principle of registration.
This is off list, so I can tell you that I am aware of many such transactions, where the use of IP address blocks change hands but where Whois does not record that fact.
Generally all that is required is for the buyer to wave a Letter of Agency at their upstream and begin advertising the block.
Yes, I have seen Geoff Huston’s slide show on multiple occasions. Were you aware that he is a proponent of needs-free transfers and the concentration of the RIRs on their primary role of Registration?
I have my own opinions on CGN as a result of deploying it as a provider and purchasing it as a user.
I have written and compiled some information at these URLs.
http://www.iptrading.com/blog-1/
http://www.iptrading.com/news/
I have read Draft-Donley and find it ludicrous. Maybe 2% of the users would even notice the move to CGN. Allowing that small slice to opt-out of CGN provides the carrier with protection against complaints of degraded service as well as at least a 10:1 reuse of addresses.
I do not fully understand the equivalence you make between the needs-test horizon (currently 24 months) and the cap on needs-free transfers for buyers who either purchase less than a /12 per year or control less than a /12 in aggregate. The goal is to reduce the paralyzing friction presented by the requirement for even tiny transactions to pass through the approval of a third party whose workings are mysterious to most buyer and sellers, in my experience. It is the cost of this uncertainty which causes sellers to eschew the slicing of their blocks into small sizes and engaging in multiple serial transactions with buyers. Likewise the costs of creating, vetting, and approving a contract are too great to induce sellers with small blocks to enter the market. My idea of allowing some transfers to happen without the uncertainty of the needs test will allow for more small transactions.
And of course would provide a disincentive for those who buy and sell address blocks without Whois updates.
And for a North American buyer of a /22 to consider opening a foreign office and receiving the addresses from RIPE or APNIC, then using them in North America?
Well, a /22 costs around $10K. I doubt that companies with needs that small would consider that option for one second. Heck the RIPE fees would eat that up in just a couple of years, not including the costs of a foreign office and the likelihood that receiving those address from RIPE or APNIC may include some attestation that they will actually be used in those geographical regions.
You keep talking about concentrating IP addresses in a few hands and the fairness of that. Remember that this situation is the status quo, and the few hands that hold the largest blocks are large companies. And remember that RIRs do not distinguish between high profit and low profit when determining need. The idea of a cap is that nobody can acquire enough space to manipulate the market, so talking about increasing concentration is not relevant. The whole point is that a cap precludes that concentration. The maximum any one entity can control or purchase per year is just .0002 of total space. That is not a recipe for establishing market control.
So I think your contention is that removing a needs test will lead to problems with hoarding and market manipulation. My answer is that limiting needs-free transfers will mitigate or ameliorate these problems.
Regards,
Mike
From: Jason Schiller
Sent: Thursday, June 13, 2013 2:23 PM
To: Mike Burns
Cc: bjones at vt.edu ; Mike Burns
Subject: Re: [arin-ppml] A Redefinition of IPv4 Need post ARINrun-out(was:Re:Against2013-4)
Off thread, trying hard not to dominate the conversation.
Mike,
I think any organization really only needs enough addresses to continue
doing IPv4 business as usual until most of the important parts of the Internet
have adopted IPv6.
(Insert appropriate values for important and most).
I expect that in some far off time, just about everything, or at least nearly all
of the things my customers want to talk to is IPv6 enabled. At this point transit
providers can safely turn up IPv6-only customers, and content providers can
start freezing their growth of IPv4 based content, growing that same content
in only IPv6.
The only people penalized at this point will be those who choose to have
legacy equipment in their own end-user networks. I think that is a
reasonable price for people to pay for refusing to move off of windows 95
or replace their DVD player. (The ball is in your court customer.)
If that far off time is 20 years, and an organization can buy 40 years worth of
address space, that is more than they will ever need, which is why I called it
"virtually unlimited". But fair point.
I don't mind large organization needing to do extra work to justify their need,
even when small ones don't.
I mind a small organization being able to secure 40 years of addressing, at
a locked in price and predictable business model, and large organizations
only able to secure two years worth, not being able to make a long term
plans to compete with smaller organizations. And the large organizations
being forced to pay the (unpredictable) market rate every two years.
"This business plan makes sense as long at the price of an IPv4 address doesn't
go above $16.52 in the next 30 years, and it doesn't goes above $12.20 in the next 5
years while we depreciate our investment. "
If we are comfortable letting organizations buy up to a cap, then that cap has to be
proportional, and represent the same amount of time.
Right now the cap is two years, and I believe the community has acknowledged we
have a problem with slow start.
I am concerned that a larger cap will lead to slowing the IPv6 deployment.
I think two years is a reasonable amount of time a large organization would need
to deploy IPv6 if they haven't started already and are serious.
Making the cap large enough for most organizations to get more than they need
to get them trough the IPv6 transition time, but limiting the largest organizations
seems unworkable to me.
Having no cap, allows the IPs to move freely to those that are willing to put up the
most money. Concentrating IPs in either rich companies, or companies that have
high value services with the greatest return on the IP address doesn't seem fair.
While I don't support this (as I believe there is already too much inequity) I support
a large (non-proportional) cap even less.
Imagine a policy of only providers of quadruple play can transfer IP addresses...
Or the two largest national incumbents in wireless, business wireline, broadband,
TV, and VoIP business sectors can transfer in up to two years of address space,
everyone else can buy as much as the can afford.
WRT forcing into CGN I wonder if you heard Geoff Huston's key note a few
NANOG meetings back:
http://www.nanog.org/meetings/abstract?id=1853
I wasn't suggesting that a /22 market cap would preclude a soft landing policy.
I was suggesting there is a /22 soft landing policy in RIPE and APNIC. That maybe a
transfer cap of /22 isn't needed as most folks who would excursive that could
just as easily go to RIPE or APNIC for their space. It is not such a burdon to have a
European or Asian office to get space.
If it is, we could do a similar soft landing proposal here.. or dove tail a soft landing
proposal like that into transfer and slow start...
__Jason
On Thu, Jun 13, 2013 at 12:35 PM, Mike Burns <mike at nationwideinc.com> wrote:
Hi Jason and thanks for your interesting reply.
As you know, my original proposal was not based on a cap that is based on the size of the entity, but instead on the number of addresses allowed to be transferred without a needs test per annum.There was a suggestion that this cap instead be placed on the aggregate number of addresses held by an organization.
The reason given was that these organizations would likely have more experience with justifications, and that the needs test would be less of a burden for them.
I agree that this aggregate cap does burden the larger organizations, but consider that these larger organizations by definition have fed at the trough of the free pool extensively compared to their less endowed competitors. One could argue that they are reaping the advantage of being there early and thus should shoulder the burden of needs testing additional transfers that late entrants who have to pay for their addresses should be exempted from.
As for me, I think either version of the cap will serve to prevent hoarding and market cornering, but will reply inline to some of your other comments:
>Consider a small rural residential ISP, with a /22.
>- This ISP is using a single /24 for loopback, point-to-point,
management network, and corporate network.
>- This IPS has 615 customers each with a single IPv4 address.
>- This ISP has seen fairly linear growth of 600 customers every two years.
>In 6 months they will exhaust their currently held space.
>They already qualify for another /22.
>Once they get this additional /22 that gives them addresses to cover 4 years.
>(/22 is about 3.4 years of customer + 6 months current available)
>A /12 represents 6,990 years worth of address space
>A /16 represents 218 years of address space
>A /20 represents 13.5 years of address space
>Should small organizations be able to by a virtually unlimited amount of
address if they can afford it?
First of all, it would be an odd rural residential ISP who could fork over $10 million for a /12.
And a /12 is not the same as “virtually unlimited”, surely you know that. The limit is just .0002 of available IPv4 space.
>Should a large organization (who can demonstrate need) only be permitted
to buy two years worth of address space?
As I said, I am agnostic about whether the cap is on aggregate holdings or annual transfers. But I would point out that it is the large companies who pose the greatest risk of hoarding addresses, simply because they usually have the required funds to risk in this endeavor.
>Organizations have also realized they only have to do native IPv4 for shortly longer
than their competitors then they can force all new customers into some sort of
provider based large scale NAT (CGN 444 + IPv6 / GCN 644).
>So now people are bracing for a slow and painful transition to IPv6.
>Never mind anti-competitive behavior of cornoring the market on IPv4 addresses,
think about reasonable players the feel the need to stockpile enough addresses
to continue doing native IPv4 longer than their competition in order to not loose
their customer base to competitors who can offer a better native IPv4 product when
you can't.
>Which means getting years worth of IPv4 space...
>Which means we are not going to run out...
>Which means we can continue to save by deferring the cost
of deploying IPv6...
>Which menas buying more space...
>(if we are not ready to deploy IPv6 buy two more years worth)
>(or if the industry hasn't embraced IPv6 in a real way buy enough to last until it has)
I agree with what you are saying, although “forcing” customers into CGN is not something I am aware of, rather this is being provided as an option to users, with the ability to opt out, at least with the British deployments and Verizon residential DSL. My Sprint 4G has been using CGN on squat space for years without my ability to opt out, though. But if you take this view, that an IPv4 market will provide a disincentive toward IPv6 deployment, maybe what you should be after is policy designed to make the market less viable, less vibrant, through increasingly onerous restrictions on transfers and allocations. Also, if ISPs can realize a 10:1 or even 100:1 address savings ratio through the use of CGN, wouldn’t they be more likely simply to introduce CGN and then hoard the addresses they have saved, rather than go to the transfer market to buy more? Note that this is only an option for those who already have large holdings of IPv4 space.
>If you are looking to make needs justification easier then maybe something like:
>- any org can transfer a single /22 no need required
>- any org can transfer up to four times the amount of address rounded to
nearest CIDR utilized in the last year
> * (jan 1, had 14 M addresses in use, dec 31 had 17M) 3M = /20 qualify for /18
>- any org who transfered a /22 can get an additional /22 when the current one is 80%
utilized even if they have utilized less than a 513 addresses in the last year
> *( jan 1 had 711 addresses, dec 31 had 820) 109 = /25
>But agin the community will have to accept a four year window, which will likely do
bad things to IPv6 deployment. If you made the threshold a /23, then you could keep
the two year window... but they why not just go to RIPE or APNIC and get a /22 from
the soft landing policy?
I am heartened that you would even consider a /22 cap for needs free transfers.
But I don’t think that this conversation should in any way preclude consideration of some kind of soft-landing policy for free pool addresses.
Regards,
Mike
Or we could let the businesses themselves decide how much they want to spend to insure themselves against a potential long-term future for IPv4. Some conservative entities may choose to buy “up to the cap” in this situation, others who think IPv6 is close, or a new transition technology is in the offing, might seek to sell addresses while they have value. It is in the interplay of those transactions that a price will be derived, representing the opinions of all transactors.
I think that what you are arguing is that a large number of buyers will have the money and inclination to buy “up to the cap”, resulting in more overall hoarding than if we maintained the needs test for all transfers. I think this would require a kind of conspiracy or at least groupthink which I do not perceive among buyers and sellers in the market.
On Thu, Jun 13, 2013 at 10:17 AM, Mike Burns <mike at nationwideinc.com> wrote:
Hi Brian,
Thanks for your thoughts.
No doubt a more vigorous transfer market will lead to more router misconfigurations.
I think a knowledgeable middle-man could help mitigate that, and would take business from the guy getting into the game without networking knowledge you mention below.
There is real uncertainty when dealing with the registries. A recent transaction took nearly a month to complete, most of which was spent in the back and forth of a justification. It’s always a fingers-crossed situation for buyer and seller. One broker told me she does the “happy dance” every time a deal makes it through justification.
Your point about moving to IPv6 is important, because that move is the 800lb gorilla in the room.
Nobody knows when the move will happen or how long it will take, but when it happens it is bound to affect IPv4 prices negatively.
Who would speculate under these conditions?
What if we limited his total purchases to a /12, or his aggregate holdings to a /12, otherwise he would be needs-tested?
Regards,
Mike
From: Brian Jones
Sent: Thursday, June 13, 2013 9:30 AM
To: Mike Burns
Cc: Mike Burns ; arin-ppml at arin.net
Subject: Re: [arin-ppml] A Redefinition of IPv4 Need post ARINrun-out(was:Re:Against2013-4)
Mike,
See inline comments.
On Wed, Jun 12, 2013 at 10:05 PM, Mike Burns <mike at nationwideinc.com> wrote:
Hi Brian,
I understand that there is a danger of overpurchasing (by whomever's definition) that comes from the removal of a needs test for transfers.
In most cases we rely on the price of the addresses to provide some check on this practice, as it would for the overpurchasing of any other asset a corporation may choose to invest in. I think we should leave those definition of what an overpurchase is to the buyers, who will have a range of intended purposes, projected growth rates, planning horizons and other considerations. At least with a cap of some sort we limit the overpurchase risk to overall address usage efficiency.
A vibrant market is one of the best mechanisms to prevent what you mention-the problem of addresses sitting idle while real need exists.
At the risk of contradicting myself, I'm not sure a vibrant market is the best answer for the networking community, but I don't disagree that what you propose would invigorate the market. See my comments below about network stability.
As the price of addresses rise and transactional roadblocks diminish, idle addresses will come into the market. As the need rises, the price will rise, driving efficiencies in the utilization of addresses and wringing the most efficiency through the highest and best use of the addresses.
I would agree that as demand rises the prices will increase, but maybe, just maybe most folks will be considering the move to IPv6 where these contentions and price increases will not exist.
And as I mentioned, due to the needs test requirement, these early IPv4 address transactions almost always involve neophyte parties on either side of the transaction, separated by language, culture, and an ocean. Often these parties are not familiar with their own RIR policy, much less the policy of another region. Most of the time the decision to sell or buy addresses has to overcome corporate inertia and antipathy to new, unusual, and unlikely-to-be-repeated transactions. This means education about the RIRs and their position squarely in the middle of the buyer and the seller.
How likely is this transaction to occur for small allocations like the /24 needed by Mr. Ryerse of this thread?
I contend that removing the needs requirement will allow for less uncertainty in what is currently a fraught process for both buyers and sellers, leading to more transactions, more price stability, and simpler transactions for all parties, including ARIN, who will avoid the time and effort of needs testing transfers.
I appreciate your contention, and it is possible that some of the things you mention may actually pan out, but I do not agree with the "less uncertainty" part of your statement. I would contend removing all needs assessment would create more uncertainty by promoting that anyone can get in the game of brokering IP addresses regardless of their knowledge about networking. Also by increasing the amount of times IP addresses get swapped around the Internet could increase the possibility for networking instability and router misconfiguration issues.
--
Brian
Regards,
Mike
----- Original Message -----
From: Brian Jones
To: Mike Burns
Cc: arin-ppml at arin.net ; Mike Burns
Sent: Wednesday, June 12, 2013 9:28 PM
Subject: Re: [arin-ppml] A Redefinition of IPv4 Need post ARINrun-out(was:Re:Against 2013-4)
Hi Mike,
I suppose it is just my old school thinking that you should be at least "this tall" to ride the ride. Given your explanations below I could relax my requirements for demonstrating technical support need for transfers. I actually didn't realize we were only considering transfers and not the remaining free blocks, so thank you for clarifying that.
It still seems that inefficient use of address space could occur when a bidder buys much larger blocks than needed due to the lack of any structured needs requirements. At a minimum a block of addresses could sit idle and unused while needs exists elsewhere. But really IPv6 should be the best solution for those needing addresses moving forward any way... :)
Brian
On Jun 12, 2013 3:15 PM, "Mike Burns" <mike at iptrading.com> wrote:
>
> Hi Brian,
>
> Thanks for your input.
>
> May I ask why you think there should be a requirement for demonstration of minimal technical need for transfers, if the reason is not to prevent hoarding and price manipulation?
>
> Remember we are talking only about transfers, and not the intelligent allocation of the remaining IPv4 free pool, and that money will be the determining factor in who receives IPv4 addresses under the current transfer policy, so long as the needs test is met. That is, we are already at a point where the highest bidder will get the addresses, irrespective of what his justified need for the addresses is, just that he has met the RIR need test.
>
> I have been operating under the assumption that the underlying reason for requiring the needs test for transfers which are already priced is to prevent a buyer without needs from damaging the market through hoarding or cornering. I understand that many people simply do not like the idea that address blocks can be bought and sold, and that money has any influence on who gets addresses, but we are beyond that now.
>
> Regards,
> Mike
>
>
> From: Brian Jones
> Sent: Wednesday, June 12, 2013 2:54 PM
> To: Mike Burns
> Cc: arin-ppml at arin.net
> Subject: Re: [arin-ppml] A Redefinition of IPv4 Need post ARIN run-out(was:Re:Against 2013-4)
>
>
> Maybe that was utopian thinking on my part. It would be nice to disregard what happens with IPv4 space but that seems to invite some sort of chaos and the last thing needed is more chaos...
>
> Intelligent allocation of the remaining IPv4 space is important in my opinion.
>
> From Dave Farmer's email earlier:
> "I think the more important issue is an appropriate criteria on the lower-end and for new enterants, the current slow-start for IPv4 isn't going to work, post-ARIN free pool. Yes, I know eliminating need alltogether eliminates that problem, but I'm not sure I can get myself all the way there. I'd like to see some minimal technical criteria that entitles someone to be able to buy up to between a /16 and a /12 and more than just that they have the money to do so. Maybe its just as simple as demonstrating efficient use of at least a /24. If you can't do that then you can only buy a /24, then you utilize it and you qualify for bigger blocks. "
>
> Regardless of whether the size blocks discussed is agreeable or not, I do agree wth the part about the need for "...minimal technical criteria that entitles someone to be able to buy up to between a /16 and a /12 and more than just that they have the money to do so."
>
> (Of course I support the idea that we all move to IPv6!) :)
>
> --
> Brian
>
>
> On Wed, Jun 12, 2013 at 11:20 AM, Mike Burns <mike at nationwideinc.com> wrote:
>>
>> Hi Brian, Matthew, and Martin,
>>
>> Can I take your plus ones to indicate support of the cap even in the face of the shell company issue?
>> (As well as support of the idea that we should all move to IPv6.)
>>
>> Regards,
>> Mike
>>
>>
>> From: Brian Jones
>> Sent: Wednesday, June 12, 2013 11:03 AM
>> To: arin-ppml at arin.net
>> Subject: Re: [arin-ppml] A Redefinition of IPv4 Need post ARIN run-out (was:Re:Against 2013-4)
>>
>>
>>
>>
>>
>> On Tue, Jun 11, 2013 at 10:42 PM, Martin Hannigan <hannigan at gmail.com> wrote:
>>>
>>> On Tue, Jun 11, 2013 at 10:24 PM, cb.list6 <cb.list6 at gmail.com> wrote:
>>>>
>>>>
>>>> On Jun 11, 2013 7:15 PM, "Matthew Kaufman" <matthew at matthew.at> wrote:
>>>> >
>>>> > When will we start caring about IPv6 and start ignoring IPv4??? Who cares if people set up shells to acquire v4 space from others? Let 'em, and get v6 deployed already.
>>>> >
>>>>
>>>> +1
>>>>
>>>> CB
>>>
>>>
>>> +1
>>>
>>> Best,
>>>
>>> -M
>>>
>>>
>>
>>
>> +1
>>
>> --
>> Brian
>>
>>
>>>
>>>
>>> _______________________________________________
>>> PPML
>>> You are receiving this message because you are subscribed to
>>> the ARIN Public Policy Mailing List (ARIN-PPML at arin.net).
>>> Unsubscribe or manage your mailing list subscription at:
>>> http://lists.arin.net/mailman/listinfo/arin-ppml
>>> Please contact info at arin.net if you experience any issues.
>>
>>
>> ________________________________
>> _______________________________________________
>> PPML
>> You are receiving this message because you are subscribed to
>> the ARIN Public Policy Mailing List (ARIN-PPML at arin.net).
>> Unsubscribe or manage your mailing list subscription at:
>> http://lists.arin.net/mailman/listinfo/arin-ppml
>> Please contact info at arin.net if you experience any issues.
>
>
_______________________________________________
PPML
You are receiving this message because you are subscribed to
the ARIN Public Policy Mailing List (ARIN-PPML at arin.net).
Unsubscribe or manage your mailing list subscription at:
http://lists.arin.net/mailman/listinfo/arin-ppml
Please contact info at arin.net if you experience any issues.
--
_______________________________________________________
Jason Schiller|NetOps|jschiller at google.com|571-266-0006
--
_______________________________________________________
Jason Schiller|NetOps|jschiller at google.com|571-266-0006
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.arin.net/pipermail/arin-ppml/attachments/20130613/c2310e14/attachment.htm>
More information about the ARIN-PPML
mailing list