On Mar 19, 2013, at 11:39 AM, McTim <dogwallah@gmail.com> wrote:

I think that perhaps we should all read the latest draft of the RFC2050 update:

https://www.ripe.net/ripe/policies/proposals/2013-03

1) Allocation Pool Management: Due to the fixed lengths of IP addresses and AS numbers, the pools from which these resources are allocated are finite. As such, allocations must be made in accordance with the operational needs of those running the networks that make use of these number resources and by taking into consideration pool limitations at the time of allocation.

McTim - Note that you are referencing an Internet Draft which is only barely announced and yet to be considered by the community. The document is actually lists several additional goals (i.e. Hierarchical Allocation and Registration Accuracy) and they are to be considered as a whole, not taken singularly: "These goals may sometimes conflict with each other or be in conflict with the interests of individual end-users, Internet service providers, or other number resource consumers. Careful analysis, judgment, and cooperation among registry system providers and consumers at all levels via community-developed policies is necessary to find appropriate compromises to facilitate Internet operations." (This is very similar to language in the existing RFC 2050.) I have no view on the policy proposal, either supporting or against, but want to make clear that the RIPE community should decide what is best for it overall and not feel constrained by a reference to just one of these goals (whether from existing or revised RFC2050) FYI, /John John Curran President and CEO ARIN

Hi, On Tue, Mar 19, 2013 at 02:43:18PM -0400, McTim wrote:

http://www.ietf.org/id/draft-housley-rfc2050bis-00.txt

I'm not sure whether RFC2050 or this personal draft is very relevant for decisions we do in RIPE community (and I've said so before). I consider the IETF to be not responsible at all anymore for the way the IANA and the RIRs handle IP address and ASN management - they have delegated the space to IANA, and accepted the existance of a bottom-up policy process for the local policies in the respective regions. RFC2050 *itself* acknowledges this in in the IESG note - it describes the *then* best current practice (in 1996). To quote from the Abstract: " This document describes the IP assignment policies currently used by the Regional Registries to implement the guidelines developed by the IANA. The guidelines and these policies are subject to revision at the direction of the IANA. The registry working group (IRE WG) will be discussing these issues and may provide advice to the IANA about possible revisions." it's not saying "the regional registries have to follow RFC2050", but vice versa, RFC2050 describes what the RIRs have been doing at that time, and explains it to the IETF community - which is good and useful, but should not be turned around. Thus, I consider any argument against this proposal solely based on RFC2050 or draft-2050-bis to be about as weak as "we have never done it this way, so we can't start now!" - times have changed, and our IPv4 policy is full of hard-to-explain historic stuff, primarily targeted at a situation that does no longer exist (LIRs justifying the size of their next allocation to the RIPE NCC). Therefore, I'd ask you to give it proper consideration, taking into account the fact that we're out of IPv4 addresses, and that this is not going to change any time soon. (This topic actually has come up at the previous two RIPE meetings when Rob Blokzijl presented his draft of an "IPv4 maintenance policy" - which didn't make it into a formal policy proposal yet, but Tore's proposal effectively does the same thing - leave in the policy document what is still relevant, throw out the rest) Gert Doering -- APWG chair -- have you enabled IPv6 on something today...? SpaceNet AG Vorstand: Sebastian v. Bomhard Joseph-Dollinger-Bogen 14 Aufsichtsratsvors.: A. Grundner-Culemann D-80807 Muenchen HRB: 136055 (AG Muenchen) Tel: +49 (89) 32356-444 USt-IdNr.: DE813185279

+1 The RIRs were always stewards of the free pool of addresses, charged with distributing them to those with demonstrated need. This stewardship was a requirement for the allocation of unpriced but scarce resources, to avoid the pitfalls of the Tragedy of the Commons. Upon effective exhaust of the free pool, the steward with the gentlest touch will step back from his stewardship and allow the market to ensure conservation. This is what markets are designed to do, allocate scarce resources efficiently. In addition to conservation, the RIRs were charged with maintaining uniqueness. In order to maintain stewardship of uniqueness, RIPE should simply act as a registrar, registering IPv4 transfers in its Whois directory. Imposing policies which restrict the free trading of addresses will endanger uniqueness as restricted trades continue underground. IPv6 is a barrier to speculators. What could be their plan, to hoard IPv4 addresses with the hope of driving up prices? That would work toward accelerating IPv6 and reducing IPv4 values. There is a natural ceiling to speculation, and there are also a billion IPv4 addresses un-advertised in the routing table. Securing enough IPv4 space to begin dictating pricing would require huge purchases from a large number of disparate sellers, and could not happen quickly or surreptitiously. I was the author of the current ARIN transfer policy, and when I introduced it, I included a limit of a /12, or about a million addresses per year which could be transferred by a single buyer without requiring a needs justification. I included this specifically as an anti-speculation measure, but the ARIN community would not allow any non-needs based transfer, and my proposal was changed to require a needs test for all transfers. This indicated to me that the real bogeyman in the ARIN community wasn't speculators, it was free markets. Is this also a RIPE bogeyman? -----Original Message----- From: Milton L Mueller Sent: Wednesday, March 20, 2013 12:16 PM To: 'address-policy-wg@ripe.net' (address-policy-wg@ripe.net) Subject: Re: [address-policy-wg] 2013-03 New Policy Proposal (No Need -Post-Depletion Reality Adjustment and Cleanup) This seems like a very well thought-out policy proposal. As the proposal points out, operational need as a criterion for free allocations only makes sense when there is a pool of unallocated addresses to conserve against excessive, unpriced occupation of number blocks. Once that basic fact no longer holds, needs assessments by the RIR serve no necessary function, but they do add bureaucracy, uncertainty, cost and the potential for arbitrariness to the process of re-allocating v4 numbers among competing uses and users. When the definition of need is based on arbitrarily defined and regularly changing time limits (3 months? 1 year? 3 years?) it illustrates how tenuous the process of needs assessment is. In fact, the concept of "need" is always contingent on price, time horizon, expected value, potential substitutes and a number of other economic factors that are best sorted out in the market. Right now, all demand for IPv4 number blocks competes with all current occupiers of the number space and all other prospective occupiers. It is competitive bidding in the market that should resolve who gets what, just as it does with labor, equipment, etc. If someone is willing to pay to extract number blocks from another holder, I don't see that the RIRs need to second-guess whether they need them or not, any more than they ask whether they actually need the amount of office space they are willing to pay for. The only argument I have heard for continued needs assessments comes what I see as unfounded fears of "speculation" and "hoarding." But these are economically driven phenomena. For these to be fatal objections, opponents of this proposal must believe that substantial acquisitions of number blocks would come from parties who have a) no operational use for the numbers and b) no interest in on-selling them efficiently to third parties who do have a use for them. Further, opponents must also believe that c) IPv6 does not constitute a viable intermediate-term option that acts as a constraint on the price of v4 transactions. I reject all three of these premises. I hope RIPE can cement its reputation as the most progressive RIR and pass this. Milton L. Mueller Professor, Syracuse University School of Information Studies Internet Governance Project http://blog.internetgovernance.org

+1 I do consider the ipv4 pool depleted and let the LIR decide how fast they want to run out themself. // Andreas Larsen Den 2013-03-20 18:50 skrev Mikael Abrahamsson <swmike@swm.pp.se>:

On Wed, 20 Mar 2013, Milton L Mueller wrote:

...

This seems like a very well thought-out policy proposal.

+1.

I consider free pool of IPv4 depleted. Let everybody sit on their allocations and use them any way they see fit. RIPE role should be to keep the database up to date and accurate. This proposal is a step in the right direction of that.

-- Mikael Abrahamsson email: swmike@swm.pp.se

Hi all, please excuse my very stumbling english. On 20/03/2013 17:16, Milton L Mueller wrote:

It is competitive biddingin the market that should resolve who gets what,

The only argument I have heard for continued needs assessments comes what I see as unfounded fears of "speculation" and "hoarding." But these are economically driven phenomena.

I think Ripe NCC has quite constantly refused to consider IP shoud be considerated as a product one could sell, and rather followed the conversation goal as a way to protect the public ressource. This is what I expect from a non-commercial structure actually, and from a regulator.

For these to be fatal objections, opponents of this proposal must believe that substantial acquisitions of number blocks would come from parties who have a) no operational use for the numbers and b) no interest in on-selling them efficiently to third parties who do have a use for them. Further, opponents must also believe that c) IPv6 does not constitute a viable intermediate-term option that acts as a constraint on the price of v4 transactions. I reject all three of these premises.

I strongly disagree with this view. First because the fact that some companies have enough money to buy high price ressource does not mean that other companies that have equally legitimate need but less ressources should be less served. Second, because letting this happen would authorize a market of abusively bought addresses develop in the hands of LIRs that so doing would significantly deviate from the Ripe goals that include to avoid waste, and not commercial abuse of a public ressource. Some examples have show already that resellers could buy an IP for 5€ and re-sell it for 10 €. That is 10,000 a /22 i.e. the space one can get buy simply becoming a LIR. This risk is particularly present while allocations reselling is now possible, which makes financial dealing of large IP spaces possible that bypass the IPv4 depletion max /22 allocation rule. I'm sorry but IPv6 does *not* yet represent a viable short-term nor middle-term option for many companies that *do* need IPv4 ressources now, either to establish their activity or to grow, wether they implement IPv6 in dual stack or not. This is a cruel evidence for some of our LIR members today. However since the changes only affect the way LIRs will make use of allocations they already have, and not deregulate the distribution of the remaining space between LIRs nor impact the allocation reselling mechanism, I am not opposed to this "do as you wish with your allocation" move. Nonetheless I feel like the conversation goal should not be completely erased from this policy, and remain. This makes sense also regarding the actual /22 max allocation rule, that *does* obey the conversation goal and should be backed by this principle. I also think the IPv4 policy should not be thought without a global vision of allocations movements and banish financial deals over public ressource. In particular, I think direct LIR to LIR reselling of empty space shoud be forbidden, and of partially occupied space sould be allowed under certain conditions aiming to garantee a business or a service is actually sold (not public ressource). Best regards, Sylvain

On 21/03/2013 10:29, Tore Anderson wrote:

1) The «public resource» is gone, so there is no longer anything to protect. This happened when the RIPE NCC reached the «last /8» on the 14th of September last year. While the RIPE NCC strictly speaking still do have space left, it is exclusively governed by the «last /8 policy», which 2013-03 does not modify.

2) While the RIPE NCC itself does not consider themselves to be in the business of "selling" IP addresses, the current IPv4 address policy *does* allow for paid LIR-to-LIR transfers of IPv4 allocations. That's the status quo, and 2013-03 does not change it one way or the other.

The public ressource is not "gone" : it was allocated or assigned, but it's still there. And since the Ripe NCC has very few in reserve, this means LIRs now have it all in their hands. I do not see the point that would explain that the "fair" distribution by Ripe NCC should be replaced by a "market-ruled" exchange by LIRs now, since LIRs are supposed to conform to Ripe policies if you relax these policies in excess then you allow LIRs to do whatever they want, whereas they could (should) still be constrained by fair use and correct documentation necessity. It seems obvious to me that LIRs that have empty allocation should not be allowed to sell it but should return it to Ripe NCC that could make it equally available to LIRs that need in on the same fair allocation logic than before. If conservation and documentation requirements are abolished then this "garbage collection" will not occur. And not making this garbage collection means a big waste. This is why i am opposed to the first two principles of this proposal, namely : « - Remove "Conservation" as a stated goal of the policy. - Removes all active policy text referring to documentation, evaluation of need, and validation of actual usage for both assignments and allocations. » Maybe this will be possible when IPv6 will be really in place which for the moment is far from being a reality, because IPv4 will not be vital for anybody anymore.

change this one way or the other. You are of course free to submit a separate policy proposal that would change this, but when it comes to 2013-03, this particular consideration seems to me to be out of scope.

Might come to this yes. Best regards, Sylvain Vallerot

Hi, On Thu, Mar 21, 2013 at 9:18 AM, Tore Anderson <tore@fud.no> wrote:

* Sylvain Vallerot

...

The public ressource is not "gone" : it was allocated or assigned, but it's still there.

The key word here is *public*. The *public* resource is gone.

not entirely, or have I missed that announcement?.

The numbers still exist in *private* pools at the LIR level and below, but the *public* pools (i.e., IANA and RIPE NCC) are empty.

"empty" in the sense that the red fuel low indicator light on my dashboard has gone on, but my vehicle still runs? I think this proposal actually lengthens the lifetime of v4, in that if an LIR makes a significant investment in v4 resources, they will be more likely to seek the longest ROI possible, thus delaying their v6 adoption. I do understand the arguments in favor, I just don't find them compelling. What is most striking to me is that we had a finite but significant supply of resources in the past, and we doled them out according to operational need. That made sense to me. Eliminate the operational needs requirement during a rationing phase just doesn't make sense to me. I also think that if adopted, this proposal would preclude an inter-RIR transfer market in that the "needs test" is required in the other regions, and that would mean that the RIPE region policies would not be "compatible" as called for in the other regions transfer policies. In other words, does 2013-03 preclude 2012-02 (if adopted) from being effective? -- Cheers, McTim "A name indicates what we seek. An address indicates where it is. A route indicates how we get there." Jon Postel

Hi,

...

I also think that if adopted, this proposal would preclude an inter-RIR transfer market in that the "needs test" is required in the other regions, and that would mean that the RIPE region policies would not be "compatible" as called for in the other regions transfer policies.

In other words, does 2013-03 preclude 2012-02 (if adopted) from being effective?

Yes and no - it depends on the policy in effect in the other region space is being transferred to or from. I'm not intimately familiar with those, but I *believe* that ARIN would be incompatible

Voluntary evaluation of need? It would be a service from the NCC. Something like: you don't need to prove need for the NCC, but if you want advice/evaluation the NCC will assist you and perform the needs based analysis. The NCC has a lot of expertise in this area. Even when showing/documenting need is not required anymore that knowledge could benefit LIRs. It could be useful for new LIRs that need a bit of guidance when assigning addresses to customers. And the NCC doing a "needs test" might make ARIN happy when doing transfers. (I just know that John will reply to this statement ;-) Cheers, Sander

Hi David,

My understanding of the intent of the ARIN's policy is NO that would not be acceptable, and such scenarios did come up during the ARIN community's discussion of Inter-RIR transfers. If it were acceptable, then someone with operational need could apply to transfer resource from the ARIN region and then transfer them to someone else without operational need once the resources were in the RIPE region, and out of ARIN's control. Rinse, repeat. If this were found to be acceptable, I highly expect there would be a policy proposal in the ARIN region to plug the leak, so to speak.

I think a realistic approach could also be to add: 'organisations that transfer IPv4 addresses to other organisations thereby show that they have no need of additional address space for 12 months. When asking the NCC for an evaluation of need the NCC will give a negative answer during this period' In other words: if you sell address space you obviously don't need it. If this WG wants 2013-03 and transfers from ARIN are the only argument against 2013-03 then it might be a solution to add this bit of (IMHO) common sense for those who want to transfer addresses from the ARIN region. But I am just suggesting options for the WG. I'll stay out of the discussion and let the community (to all who are reading this: that includes you!) decide :-) Cheers, Sander

In other words: if you sell address space you obviously don't need it.

False, as a blanket assumption. Obvious counter-scenario: You need a /16 but hold a /8. You sell the entire /8 due to the premium it commends as a contiguous block but then buy a /16.

If this WG wants 2013-03 and transfers from ARIN are the only argument against 2013-03 then it might be a solution to add this bit of (IMHO) common sense for those who want to transfer addresses from the ARIN region.

But I am just suggesting options for the WG. I'll stay out of the discussion and let the community (to all who are reading this: that includes you!) decide :-)

Cheers, Sander

On Mar 21, 2013, at 6:30 PM, Sander Steffann <sander@steffann.nl> wrote:

Voluntary evaluation of need? It would be a service from the NCC. Something like: you don't need to prove need for the NCC, but if you want advice/evaluation the NCC will assist you and perform the needs based analysis. The NCC has a lot of expertise in this area. Even when showing/documenting need is not required anymore that knowledge could benefit LIRs. It could be useful for new LIRs that need a bit of guidance when assigning addresses to customers.

And the NCC doing a "needs test" might make ARIN happy when doing transfers. (I just know that John will reply to this statement ;-)

ARIN's Inter-RIR transfer policy can only happen to an RIR which shares a "reciprocal, compatible, needs-based policy". A voluntary mechanism does not equate to a "needs-based" policy, so we would have a conflict. As noted by others on this list, there is nothing fundamentally wrong about such a conflict, other than the fact that it would mean additional discussion in both regions about the value of alignment and options for getting there. FYI, /John John Curran President and CEO ARIN

Hi Sylvain, * Sylvain Vallerot

But they can also reveive much bigger spaces from buying them. Then they will no be constrained anymore by the fair use policy.

So what could happen is that large IP spaces get sold and used thus exceeding very much the simple /22 spaces you mention. Maybe sometimes "fair use" users could even be moved away to make spaces available for selling in the worst cases (I know this is a very pessimistic example but I'm pretty sure it will happen). Less dramatically never decreasing cases of "for money" or simply "bad" assignments will occur, just like entropy does (quantity of mess in a natural system).

Yes, under the proposed policy it would be technically possible for an LIR to obtain (through transfers) a larger amount of address space than they have an operational need for. However, is this going to be a problem? Realistically? With pretty much anything here in life, it is possible for an individual or an organisation to buy more of something than he need. I can go to the store and buy all the bread there is, even though I'd be full after a couple of slices. My employer could rent all the office space in the building I'm sitting in, even though everyone fits comfortably in two floors. And so on, and so on... In the same vein, yes indeed - an LIR could, in theory, obtain more address space than they have an operational need for. There's a perfectly logical reason why this doesn't happen, though. Individuals and organisations simply have better things to spend their money on than stuff they don't need. Not wasting money on useless crap is economy 101 - money is a scarce recourse, and there's always going to be something useful you can spend it on instead. (This is much like IPv4 addresses nowadays, by the way, which is why I expect 2013-03 will not cause LIRs to instantly assign away all their remaining space.) So to be honest, I don't find the argument that some LIRs will just go and buy lots of IPv4 address space they don't need "just because they can" particularly sound. Common sense tells me that if they have no actual need for IPv4 addresses, they won't be spending money on them either. However, let's for the sake of the argument assume that such LIRs do exist. The space they're obtaining from another LIR is necessarily unused to begin with (that's a current precondition for a transfer to be approved which is kept by 2013-03). So instead of having LIR "A" (the seller) sit on the space and not using it, the only actual change is that now LIR "B" (the buyer) is doing the exact same thing. The space is still just as unused as it was before the transfer. So. What is the damage done to you, I, and the rest of the RIPE community and membership? I honestly don't see any.

I believe a good policy should ensure free space returns to available pool for fair use needs, but since allocation transfer is allowed it should not be an allowed bypass of such a fair policy.

I understand you have this point of view, however please realise: It is not germane to the discussion of 2013-03, because the status *today* is: - LIRs are not obligated to return free space to the available pool. - LIRs may to transfer allocations directly to other LIRs. Proposal 2013-03 *does not change these facts in any way whatsoever*. Like I said earlier, if you want to see the policy changed so that direct LIR-to-LIR transfers become prohibited, then you are free to submit a policy proposal that does this. It should be relatively easy, I think (start by deleting the entire «Transfers of Allocations» section). Then we can discuss that proposal on its own merits, in a separate thread. Please leave it out of the 2013-03 discussion though, as it is irrelevant here.

With time, users and LIRs will just stop using IPv4, but until this happens I believe we need fair use and conservation principle.

NB: "private" space does not exist for me : internet *is* public, all of it, and ressources are delegated to LIRs, then users, but not owned by them.

There is a major difference between public (IANA, RIR) and private (LIRs) pools. The public ones are available for all to allocate from in an equal and fair manner. The NCC is simply not at liberty to refuse to make an allocation that was valid according to the policy. The LIRs, on the other hand, are under no obligation to assign address space to anyone. So while you are of course free to call the LIR pools "public" if you want, know that the public have no access to them. Best regards, -- Tore Anderson

* Jim Reid

I fear we're in danger of rat-holing and trying to over-engineer 2013-03. It's all very well for the proposal to enumerate every hypothetical scenario and threat. But what good does that actually do? An in-depth discussion of this proposal is like having a debate on the Titanic some time after the ship has hit the iceberg about which wines would be the best match for our dinner choices.

Indeed. My goal with 2013-03 to reduce the amount of bureaucracy needed to operate an LIR (and by extension, the RIR), and reduce the amount of defunct policy text. To make it neat, relevant, and concise. If I had tried to instead close every tiny loophole and theoretical potential for abuse, however unlikely, the opposite would have been the result. I might as well have asked a law firm to write it.

The whole point is our policy-making machinery should be responsive and quick. So if 2013-03 later turns out to be unsatisfactory -- for some definition of unsatisfactory -- we change or replace it in light of the information available at that time.

Yes. And with regards to the particular concern over an LIR without operational need attempting to corner the market by buying everything there is, I'd like to ask the WG to keep in mind that there is a 24 month cooling off period for transfers which is retained by 2013-03: «LIRs that receive a re-allocation from another LIR cannot re-allocate complete or partial blocks of the same address space to another LIR within 24 months of receiving the re-allocation.» If it does turn out to be an attempt to corner the market after 2013-03 passes, we do have sufficient time to take corrective action (through new policy) against the "cornerer" before he has a chance to cash out. This in itself is a deterrence for anyone seriously considering an attempt to corner the market. So is the possibility that IPv6 deployment will increase during those 24 months and lower our LIRs' interest in, and willingness to pay for, IPv4.

BTW, common sense flies out the window whenever there is a (perceived) market scarcity. [Back in the 1970s there was panic buying of salt in the UK because of a word of mouth rumour about "the workers going on strike in the Siberian salt mines".] If there's the equivalent of panic buying of the last dregs of v4, so what? Once it's gone, it's gone and then we can all get on with what we should be doing: deploying IPv6. No amount of policy making today is going to prevent panic buying or other irrational behaviour by those who see (or think they see) empty IPv4 shelves in the shops.

If LIRs have operational need, they can engage in panic buying under the current policy too. Presumably, those that participated in the panic buying of salt you refer to were folks that actually needed and used salt... I don't see why anyone who has no need for X would get panicky about a scarcity of X, to be honest. Whether X equals IPv4 allocations, or salt. But that's just me and my (not so) common sense, I guess. ;-) Best regards, -- Tore Anderson

On 25/03/2013 10:22, Tore Anderson wrote:

Yes, under the proposed policy it would be technically possible for an LIR to obtain (through transfers) a larger amount of address space than they have an operational need for.

And assign it without any control nor fair use because of the removal of the conservation goal and means (documentation, Ripe control). Of course LIRs won't just buy to waste money. But some would to sell rare ressource (a lot already do).

With pretty much anything here in life, it is possible for an individual or an organisation to buy more of something than he need. I can go to the store and buy all the bread there is, even though I'd be full after a couple of slices.

It is possible but not necessarily worth it. It is not worth buying all bread if everybody can get it for low price and it is available. It is not worth buying it all either if you cannot resell it to whom you want. IPv4 ressource is rare so first condition is met. Selling to richest first is possible with 2013-3 deregulation. What happens when a vital ressource becomes rare usually ? Free market is regulated, rationing is set in place for everyone to get a chance to survive. Cheaters make black market that restores "free market" logic, where richest buy first and poors starve. Allocation transfer allows the black market to I does not lead to abuse since regulation of final step selling is in place. Deregulation makes money oriented reselling allowed. This is black market made legal, rationing screwing. I think you got my point. Best regards, Sylvain

On Tue, 26 Mar 2013, Tore Anderson wrote:

So imagine you're an LIR. You've just spent 1 million euros on a good-as-new /16. What on Earth would cause you to then go ahead and «assign it without any control»?

Exactly. There is no need to push conservation for IPv4 from the outside anymore. RIPE and other RIRs should keep the documentation up to date and let people trade their addresses and use them as they see fit. IPv4 is exhausted. The stone is bled try. Stop trying to squeeze it. Let people who want to stay in IPv4 land pay for the privilege. There is IPv6 addresses at low-low prices, let the people who go that route pay for some of the (potentially higher) investment by either selling, or not having to buy IPv4 addresses. PS. I personally believe the prices some brokers are trying to sell addresses are way too high. If the price was USD 5-20 per IPv4 address (as they seem to think), I believe there will be a lot of sellers. I have yet to hear someone wanting to *buy* at those price points, only sellers. Those price points would pay for a lot of CGN boxes and still make a hefty profit. -- Mikael Abrahamsson email: swmike@swm.pp.se

* boggits

On 26 March 2013 08:47, Mikael Abrahamsson <swmike@swm.pp.se> wrote:

...

On Tue, 26 Mar 2013, Tore Anderson wrote:

...

So imagine you're an LIR. You've just spent 1 million euros on a good-as-new /16. What on Earth would cause you to then go ahead and «assign it without any control»?

Exactly.

Because I can get 4 million by breaking it into smaller chunks at a higher per unit cost...

The LIR holding the original allocation could do the exact same thing, so what is the damage inflicted on the RIPE community and membership here? IMHO it would is much more likely that a original LIR would do something like this, rather than a "middleman LIR", because such a middleman LIR will need to wait 24 months before it can start selling the smaller chunks on to other LIRs. -- Tore Anderson

On 3/25/13 9:16 PM, Sylvain Vallerot wrote:

Selling to richest first is possible with 2013-3 deregulation.

Welcome to the world of basic business rules :) Cheers, Jan P.S: Natural selection mechanism works for us for a very long time already ;)

On Mar 25, 2013, at 2:09 PM, Milton L Mueller <mueller@syr.edu> wrote:

... Do you seriously think there is a major likelihood of lots of people making winning bids for number blocks just for the heck of it? Given the incredible flexibility and even subjectivity of the concept of "operational need" we are adding massive bureaucratic costs and delays but the gain achieved is completely unclear to me.

I note that the secondary market for radio spectrum resources in the U.S. and elsewhere does not require acquirers to prove to anyone that they "need" it, yet there is no big problem with the way it has worked.

Just to complete your analogy... If RIPE were to operate with policies similar to the US radio spectrum market, it is possible that the absence of needs-assessment would not automatically lead to rampant speculation. Of course, this is also because to perform a transfer in the US spectrum market, you're likely going to hire lawyers, complete a 40 page form (FCC Form 603) and "certify that the proposed transaction meets specific criteria indicating the absence of potential pubic interest concerns relating to eligibility, use restrictions, foreign ownership, designated entity policies, and competition." <http://hraunfoss.fcc.gov/edocs_public/attachmatch/DA-05-2226A1.pdf> Postulating what might or might not happen with a transfer market in the RIPE region (in the absence of needs-assessment) based on history with the US radio spectrum market is disingenuous without also noting the fact that the US radio spectrum market has its own "bureaucratic costs and delays" of a completely different nature which may easily impact the resulting dynamics. FYI, /John John Curran President and CEO ARIN

* McTim

I think that perhaps we should all read the latest draft of the RFC2050 update:

https://www.ripe.net/ripe/policies/proposals/2013-03

1) Allocation Pool Management: Due to the fixed lengths of IP addresses and AS numbers, the pools from which these resources are allocated are finite. As such, allocations must be made in accordance with the operational needs of those running the networks that make use of these number resources and by taking into consideration pool limitations at the time of allocation.

This proposal seems counter to the above, as well as in conflict with 2050 itself.

I note the author has tried to provide counter arguments, but to me they are not sufficient to persuade me to support this proposal.

Hi McTim, The pool from which all allocations are drawn in the RIPE region is exclusively governed by the «last /8 policy», currently found in section 5.6 of the policy document. In a nutshell, it says: «a single /22 per LIR, no more, no less - regardless of actual operational needs». 2013-03 does not change the mechanics of this last /8 policy, at least that is not my intention. The proposal does "promote" it to be the main and only allocation policy found in the document, instead of being contained in section 5.6 all by itself - but again, the intention of this change is just editorial cleanup, clarification, and getting rid of defunct policy text - not to actually modify the effective policy. So, in summary, I don't see 2013-03 running any more afoul of the 2050-bis passage you quoted than what the current last /8 policy already have. Best regards, Tore Anderson

* James Blessing

Whilst the address being allocated to LIRs could follow this model the assignment of address space to end users should continue on a needs basis.

Hi James, For what it's worth, my opinion is that it will indeed continue in this manner, even if 2013-03 is accepted. The way I see it, this will follow naturally from the depletion state we're in. LIRs know by now that they can no longer come running to the NCC to refill their pools should they run empty, hence they have all the incentive in the world to conserve what they have already been allocated, and therefore only assign to its End Users the space that the End User actually need and can justify. 2013-03 leaves it to the individual LIR to decide on what periods and criteria makes the most sense in each case, though. Should a LIR on the other hand opt not to do so, and assign to its End User the amount of addresses the End User asked for with no limitation, that LIR will realise soon enough that it has done a grave mistake. However, that mistake only hurts that LIR only - you, me, and the rest of the community are not affected by that LIRs mistakes (unlike what we would be if the same thing happened pre-depletion). The upside is that under 2013-03, if you or I get a new customer signing a, say, three-year contract, that leaves no doubt that he has a valid need for addresses in that period - the policy will actually allow us to make that assignment. Best regards, Tore Anderson

* Sebastian Wiesinger

For us I don't see a big problem as I can implement an internal set of rules regulating the assignment of our remaining space. It would help, however, if the RIPE NCC would publish some sort of guidance document on how to establish/document the need for IP addresses. I assume that the RIPE NCC already has something like that for their resource analysts?

Hi, Existing LIRs are supposed to have something like this in place already. When an LIR is making an assignment that falls within its assignment window, that LIR alone is responsible for evaluating their End User's need and supporting documentation (and archiving it). Should 2013-03 pass, LIRs are free to continue justifying their assignments using the exact same criteria that they did before. The ripe-583 request form will formally be obsoleted by 2013-03, but that doesn't mean that LIRs can't continue to use it as if nothing happened. Tore

Hi Alex, * Alex Le Heux

This proposal is a gigantic leap in the right direction.

Thank you. I hope that the WG chairs and I can interpret this as a voice of support, despite what you write about PI below?

Any future IPv4 policy should describe what to do with PI assignments [...] Leaving them in limbo is not a good idea.

I want to avoid having my policy proposal fix "everything" at the same time. While it may be tempting to have it fix issues A, B, C, D, E, F, and G in one go, the risk with that is that the community doesn't agree with, say, F - and on that basis alone killing the entire proposal. Therefore I want to limit the scope of changes as much as reasonably possible, so that we can discuss the actual issues only without getting side-tracked over another "nice to have" change that might prove controversial. For what it's worth my primary motivation for submitting the proposal (i.e, the "A" issue, or the itch I'm actually trying to scratch) is to get rid of the need justification bureaucracy relating to making PA assignments. In a nutshell: I know what my customers need for the contract periods they sign up for, and I want to be able to assign that space to them without being constrained by RIPE policies and without having to fill out any RIPE paperwork. The "B" issue it tackles, is getting rid of the need justification bureaucracy for [transfers of] allocations. This is not the main goal, but due to A it would make absolutely no sense to leave it in - otherwise any LIR could just synthesise whatever allocation need they wanted to (without actually running afoul of the policy). The "C" issue fixed by my proposal, the clean-up and restructuring, is due to A and B being so deeply intertwined in so many parts of the policy. I first tried to «surgically remove» need justification text from all parts of the policy in order to solve A and B, and leave everything else intact, but I soon realised that it was easier to just delete all the old and obsolete sections completely, and (if necessary) replacing them with whatever the currently active policy (i.e., the "last /8" policy) had to say.

The cleanest, and most radical, solution would of course be to say "addresses are addresses" and do away with the difference between PI and PA and strip all references the "PA" and "PI" from the policy and RIPE DB.

I've heard talk about such a «PA/PI unification» proposal, at least for IPv6, but I don't think anything was ever actually submitted to the PDP. I would prefer that such a change could be proposed and discussed independently of 2013-03, though. That said: The eventual passing of 2013-03 would be a great benefit to such a proposal, as the amount of policy text that needs to be changed would decrease dramatically. See also: http://www.ripe.net/ripe/mail/archives/address-policy-wg/2011-October/006496... Best regards, Tore Anderson

Hi Alex, * Alex Le Heux

But, by ignoring PI and cutting it out completely, I do feel that your proposal creates a limbo where there was none before.

Questions that will remain unanswered under the proposed policy include, but are probably not limited to:

As I understand it, if something is not explicitly mentioned in the policy, it is by default not permitted. With that in mind, I'll try to answer your points below:

- Is a PI assignment valid as long as the original criteria remain valid?

Yes, my proposal does not change or remove the relevant text that states this. See section 6.3.

- What about transfers of PI? Not allowed? Allowed?

Not allowed, because no explicit policy that allows it exists. Note that this is also the case in current policy, so my proposal upholds the status quo.

- Are sub-assignments now allowed?

Not allowed, because no explicit policy that allows it exists. Note that this is also the case in current policy, so my proposal upholds the status quo.

- Is there a requirement to keep the registration data in the RIPE DB up to date?

Yes, my proposal does not change or remove the relevant text that states this. See sections 3.0 (point 3), 4.0, and 6.3.

Even though the RIPE NCC will most likely not assign new PI in the future, these are still issues that its staff will have to deal with, and by extension, we as a community as well.

The way I see it, ASSIGNED PI is now simply a historic status, in the exact same way that, e.g., ALLOCATED PI and ASSIGNED ANYCAST is. And it remains documented as such in section 7.0 of the proposed policy. I do not quite see why ASSIGNED PI specifically needs more documentation than those other historic types of delegations. That said, if you can suggest some text to add that would resolve your concerns, I would be very happy to incorporate it into the proposal. Tore

Hi, On Mon, Mar 25, 2013 at 01:59:47PM +0100, Daniel Stolpe wrote:

[ PA/PI unification ] Yes. Mayby we have what we have with IPv4 but I support Gert in the attempts of bringing som order to the IPv6 space.

Yeah. For the record: this has not been forgotten, but again I have been a bit too busy to make a formal policy proposal (with specific text, etc) out of my ideas. I think I'll try to round up a few volunteers in Dublin to help me with this... Gert Doering -- APWG chair -- have you enabled IPv6 on something today...? SpaceNet AG Vorstand: Sebastian v. Bomhard Joseph-Dollinger-Bogen 14 Aufsichtsratsvors.: A. Grundner-Culemann D-80807 Muenchen HRB: 136055 (AG Muenchen) Tel: +49 (89) 32356-444 USt-IdNr.: DE813185279

On Tue, Mar 19, 2013 at 5:00 PM, Emilio Madaio <emadaio@ripe.net> wrote:

Dear Colleagues

A proposed change to RIPE Document ripe-582, "IPv4 Address Allocation and Assignment Policies for the RIPE NCC Service Region", is now available for discussion.

You can find the full proposal at:

https://www.ripe.net/ripe/policies/proposals/2013-03

We encourage you to review this proposal and send your comments to <address-policy-wg@ripe.net> before 16 April 2013.

supported -- Kind regards, --- Dimitri I. Sidelnikov FREEnet Hostmaster

On Mon, 25 Mar 2013, Jan Zorz @ go6.si wrote: I support this proposal.

On 3/19/13 5:00 PM, Emilio Madaio wrote:

...

Dear Colleagues

A proposed change to RIPE Document ripe-582, "IPv4 Address Allocation and Assignment Policies for the RIPE NCC Service Region", is now available for discussion.

You can find the full proposal at:

https://www.ripe.net/ripe/policies/proposals/2013-03

Support.

Way too much energy is currently wasted in battling for nickels on the bottom of the barrel.

Cheers, Jan

-- Mvh Fredrik Widell Resilans AB http://www.resilans.se/ mail: info@resilans.se , fredrik@resilans.se phone: +46 8 688 11 82

* Richard Hartmann

I support this proposal.

Thank you!

Three cosmetic comments; it would be nice if they could be incorporated. If not, they should _not_ hold up this process in any way.

* "In the event that this /16 remains unused at the time the remaining addresses covered by this policy has been distributed, it returns to the pool to be distributed as per section 5.1, and this section is to be automatically deleted from the policy document." should read "have been distributed", instead.

No-brainer. Will incorporate, thanks.

* "IXPs holding other PI IPv4 space for their peering LAN (i.e. they are seeking a larger assignment), must return their old peering LAN resources back to this pool within 180 days of assignment." uses IXP once and Internet Exchange [Pp]oint otherwise. Might clean that up just as well. Similarly, "Internet Exchange point" should be replaced with "Internet Exchange Point"

I suggest abbreviating all references except for the first one (where the abbreviation is defined in the first place). Sounds good?

* "Clear contractual arrangements are recommended and are mandatory for PA space." This could be worded more clearly. Are they recommended generally, but mandatory for PA? If yes, PI have contractual requirements as well so this could be mentioned here as well. An alternative interpretation would be that contracts are both mandatory and recommended for PA space which would be somewhat nonsensical.

The original statement read: «LIRs must make it clear to End Users which type of address space [PI vs PA] is assigned. Clear contractual arrangements are recommended and are mandatory for PA space.» I removed the first sentence as part of "cleanup", since PI is a thing of the past. Agreed that the remaining statement looks awkward on its own. Suggested new replacement: «Clear contractual arrangements are mandatory for PA space.» In any case, contractual arrangements have been mandatory for both PI and PA space since 2007-01 was implemented, so the suggestion that it was "only" recommended for PI was in any case obsolete and incorrect policy. Good catch, thanks! Tore

* Richard Hartmann

I still hope that IPv4 PI will come back, but that's for another day.

Yep, that's 2012-04. While 2013-03 will remove lots of *currently* obsolete text regarding PI, there's no reason why 2012-04 can't add the now-relevant-again bits back in.

...

«Clear contractual arrangements are mandatory for PA space.»

How about simply extending it to cover everything:

Clear contractual arrangements between LIRs and end users are mandatory for all RIPE resources (IP space and AS numbers).

No. This policy document changed by 2013-03 covers IPv4 address space only, so ASNs, IPv6, etc. is out of scope. Also, I don't want to have the proposal "accidentally" change the effective policy wrt. what kind of contractual arrangements are mandatory at what time, and since there's many types of legacy delegation types in existence there's a non-zero chance that removing the explicit reference to PA space in the above sentence will cause "jurisdiction creep", which in turn may be controversial in its own right, and require me to spin another version, and so on. I'd rather avoid taking that chance at this point. In a nutshell: 2013-03 amputates the dead policy limbs. Cosmetic surgery can follow later. :-) Tore

Hi, Tore -- On 30/03/2013 14:01, "Tore Anderson" <tore@fud.no> wrote:

In a nutshell: 2013-03 amputates the dead policy limbs. Cosmetic surgery can follow later. :-)

Cleaning stuff up is good, and this policy is a great step - thanks for bravely taking up the challenge. I support the ambition of the policy, and happy to support the current wording. Richard's English tweaks should be included as well. There was a lot of debate about whether to remove that 'conservation is an aim of the policy' at the start - I would support a 2013-03-bis which does not delete the conservation line if this is required to build consensus, but it's time to delete the legacy stuff. Andy

* Gert Doering

Actually, for *PA*, they have always been "recommended" - mostly to ensure that customers are aware that they can't normally take the PA space away to other providers.

For PI and AS numbers, they are mandatory since 2007-01.

So changing this to "mandatory" for PA space would, in my reading, be a change of policy (... and opens up a new opportunity for bureaucracy, as in "being required to display the contract to the RIPE NCC upon an audit" or such).

Hi Gert, The currently active policy document (ripe-582) says: «LIRs must make it clear to End Users which type of address space is assigned. Clear contractual arrangements are recommended and are mandatory for PA space.» While I'm not a native English speaker, I have severe difficulties interpreting the above statement as anything but: * PA space -> clear contractual arrangements are *mandatory*. * Non-PA space -> clear contractual arrangements are *recommended*. This is from a section titled «PA vs. PI Address Space», so the way I interpret it it, you may substitute "Non-PA" with "PI" above, ergo we're left with: * PA space -> clear contractual arrangements are *mandatory*. * PI space -> clear contractual arrangements are *recommended*. Clearly, the last point is in conflict with 2007-01, but I'd assume that is just an oversight - 2007-01 should have updated it, but did not. Assuming 2007-01 takes precedence, then, the actual policy is: * PA space -> clear contractual arrangements are *mandatory*. * PI space -> clear contractual arrangements are *mandatory*. I hope you can follow my interpretation, or if not at least can point out where my mistake is. It it definitively *not* a goal of 2013-03 to add any contractual requirements where there was previously none, but it seems to me that in the PA case, they are already there. In this regard it we might consider it a goal of 2013-03 to not change anything, one way or the other (because that'd be "scope creep" which I really want to avoid). Tore