2007-08: response to ETNO position
Dear all, and in particular the representatives of ETNO, Let me first thank ETNO for taking the effort, twice, to make a valuable contribution to the discussion surrounding the 'Afterlife' of IPv4 in the form of position papers. It is good to see that we all seem to agree on the following fundamentals: - The available free pool of "allocatable" IPv4 addresses will run out - quite likely by 2012. - The existing IPv4 network will continue to operate for many decades to come. - There will still be demand for IPv4 addresses. - After the available pool of IPv4 addresses is exhausted, needs (even limited) will continue to appear that are provided with other options - The only long-term solution to the "unavailability" of IPv4 addresses is the widespread adoption and deployment of IPv6 infrastructure, transport and client services. (This is pretty much a literal copy from the ETNO position; I could have used my presentations on the subject as a source as well) The other point that most of us will readily agree with, is that we don't like markets. Markets for IPv4, of course. I'm not particularly fond of the idea, either. I would encourage anyone who is completely in favor of an IPv4 market to establish to visit the next ICANN meeting and see first hand what a thriving market has arisen in the domain name space. But there the agreement stops. Let me first rephrase what 2007-08 is about, before responding to ETNO's position. 2007-08 is about setting up a framework for (IPv4) resource transfers between LIRs. Nothing more, nothing less. No IPv6, no ASNs. No PI, no ERX. It is a framework in the sense that, in my opinion, it encompasses the absolute minimum set of policy changes required for this to happen without immediately shooting ourselves in the foot. It is a starting point, not an end point for transfer policies. It does not create a market. We don't stop making policies right after passing this one; the proposed framework can be adapted, altered, even abandoned if we see as a community see fit. The point about moving on urgently with 2007-08 is that it does make a significant impact on how the RIRs run their shop. Implementing and testing 2007-08 takes time, probably a lot of time. Having it in place before the RIRs run out means that we at least have policy we can execute when we run out. Current policy would even allow a single request for a large block of addresses to completely clean out the reserves a RIR might still have in smaller available blocks. But I digress. ETNO, in its position paper, provides a number of principles that a 'post-IPv4' (ETNO's term, not mine) world should obey. I'll mention them shortly and then reply. 1) Identification of legitimate use ETNO would like, in short, to be able to identify parties that have genuine rights of use over IPv4 address prefixes. It is also argued that this can only be done globally, not regionally because that 'will be unworkable'. This global approach must also be in place long before the period in which IPv4 is harder to get. While I agree that we need to keep track on who has rights on what - it is one of the key points of 2007-08 - doing it globally would mean abandoning all RIR databases for a global registry. And this global registry is required as of today. As you all know, IPv4 is already harder to get; policy has changed last year so LIRs can only get enough for a 1 year period rather than a 2 year period. In my opinion, this is a 'pie in the sky' idea. It is unlikely for a global policy to establish itself within any reasonable time if that policy would abandon the single most significant asset of any RIR; their database. 2) Transparency Any decision regarding allocation or assignment must be done in a transparent and public way and consistent with the bottom-up policy process. Any deficiencies should be removed from policy before then. And globally, of course. I fully agree on transparency. What I fail to see is what is meant with 'public way' and why 2007-08 would fall short of that mark. The RIRs don't currently make the arguments based on which an LIR receives an allocation public, either. When IPv4 space moves hands because of a merger or acquisition this doesn't happen either. That's all current bottom-up policy. 3) Fair and neutral reuse All recovered address space should be recycled in a globally synchronized fashion. RIRs should work harder recovering IPv4 space. I think that this is a good idea, but let's not overestimate what IANA and the RIRs are able to recover. Current global demand is among the lines of 13 /8s a year. Taking low-hanging fruit that has already been recovered into account we can perhaps recycle another dozen /8s. Recovering those will take years. The same goes for re-tasking any experimental or reserved space to 'global unicast'. It would take years to accomplish and only have a very limited impact. As I've said in one of my presentations, even if we were to find a brand new and completely empty IPv4 space somewhere (which will not happen) we'd probably use that up in a decade as well. The fundamental issue is consumption, not lack of production. 4) Self-regulation Allocation of IPv4 addresses should always be guided and regulated by the bottom-up process of the RIRs. Fully agreed. What we probably do not agree upon is what comes out of this process. In addition, one of my main reasons for pushing 2007-08 is that I would very much like that the RIRs stay relevant and the process stays in place. Removing IPv4 from the grasp of the RIRs (and therefore also from the bottom-up policy process) is the single most significant risk of not facing the 'afterlife' in a timely fashion. The moment the RIRs say 'no you can't have more space and you're not allowed to get it anywhere else' is the moment the world turns to another place for that resource. I find it curious that ETNO puts significant effort into writing position papers such as this one and not join us in that bottom-up process, but that's a different matter. 5) No connection between IPv4 and IPv6 Policies we make for the 'afterlife' of IPv4 should not affect any policy with regard to IPv6. Yes. 2007-08 is in compliance with that. In another chapter, ETNO also addresses a few aspects they think should be taken under consideration. In their point of view, we can either get a reclaim/reuse model or a transfer model. I think we can get both, provided we change allocation policy so that a single request can not wipe out carefully built-up reserves. This falls outside of 2007-08 or any other current policy proposal but I'd expect a proposal sooner rather than later. In describing transfers, ETNO argues that transfers are the first step towards a market. That is false. It is the first step towards a market under the control of the RIRs and the current bottom-up policy process we all know and love. A market will evolve whether we accommodate it or not. The report also reiterates that by adopting transfers we abandon that bottom-up policy process. I don't remember putting that in my proposal - perhaps I should read it again. To finish up, the ETNO paper worries about the options RIRs might have to enforce policies in a transfer situation. I'll let you in on a secret: how many times has a conflict been brought to the RIPE NCC arbitration council? Once. Just one. That either means that we all gladly and completely abide set policy or that the real world options RIRs have to enforce policy are limited in the first place. I'll let you choose. In summary: Thanks for bearing with me - this is quite a bit longer than I anticipated it to be. The ETNO position paper gives us a fascinating view of the 'afterlife' and tells us that none of the options we currently have will be an acceptable substitute for the IANA free pool from the ETNO perspective. It also tells us that we should abandon regional policy while at the same time keeping the current bottom-up policy process. This puzzles me. The one reason I could find in the entire paper why ETNO does not like 2007-08 is that it does not like the idea of a market. Neither do I. But 2007-08 is not about a market. It is a fallacy to think that any of us can either create a market or prevent one from being created. That is particularly so in an environment that will have global demand and (potentially) supply. 2007-08 is about staying relevant in an environment where a market evolves. I am fully aware that an evolving market will bring all sorts of interest, not in the last place from regulators. But it's not the regulators we should fear; letting it all slip through our hands is what we should worry about. Creating a transfer policy shows that we as a community can be part of a solution and show good stewardship, not just part (and cause) of the problem. People who are part of the solution will have a seat at the table - the people who merely caused the problem will not. Inaction is not a sign of good stewardship. So we need to be seen doing something, preferably something that will make a difference. So let's consider the alternatives: either we enable transfers or we prevent running out of IPv4 space. I can come up with a policy proposal for that as well - I'm positive you're going to like it even less than this one. If we would only allow justification for new v4 allocations to be based on servers and network infrastructure and not on 'eyeballs' we'd see our consumption rate evaporate overnight and the IANA free pool will last us at least a decade longer, while at the same time forcing the 'eyeballs' to move to IPv6. Think that's a better idea? Personally, I think not. I would urge ETNO to become part of this discussion and work with the community in our prized bottom-up policy process, rather than setting limits to what can and can not be discussed an not coming up with solutions themselves. I thank you for your time and look forward to your responses. Best, Remco van Mook (no hats)
Let me first rephrase what 2007-08 is about, before responding to ETNO's position. 2007-08 is about setting up a framework for (IPv4) resource transfers between LIRs. Nothing more, nothing less. No IPv6, no ASNs. No PI, no ERX. It is a framework in the sense that, in my opinion, it encompasses the absolute minimum set of policy changes required for this to happen without immediately shooting ourselves in the foot. It is a starting point, not an end point for transfer policies.
However, we already have a framework for IP address transfers between LIRs which works substantially the same in all RIR regions. If it involves a change in corporate ownership, you simply need to show proof of the change and that the network is still operating under the new entity. The RIR then updates all records showing the new ownership. Or you give the address block back to the RIR because you no longer need it, and some other LIR will get those addresses at some point in the future. This framework works. There is no reason to believe that it would cease working as IPv6 deployment picks up. Most network operators are undertaking some form of audit of their internal use of IP addresses in preparation for the shortage, which means that as IPv6 picks up, they should know when they have surplus IPv4 addresses and can return them to RIPE or whatever RIR they came from. If it is not broken, then why would we want to change this framework? You claim that this will speed up the process of transfers but I cannot see how adding a new set of transfer rules will speed up anything. Even though you claim that this is an LIR-to-LIR transfer without RIPE involvement, this is not true since there are various rules which RIPE must administer and RIPE can disallow a transfer if, for instance, an LIR has received a re-allocation within the past 24 months.
1) Identification of legitimate use
While I agree that we need to keep track on who has rights on what - it is one of the key points of 2007-08 - doing it globally would mean abandoning all RIR databases for a global registry.
Huh? I can't see where you get this idea. This ETNO point is in full support of the existing RIR system, where every IP address range is under the authority of one of the RIRs, and where the identity of the "legitimate user" can be authoritatively discovered in the RIR's database (or whois directory). In this point ETNO is supporting the existing transfer framework in which all IP address transfers must go through one of the 5 RIRs. Under 2007-08 it is possible that some LIRs will transfer addresses which an RIR will not register in the database. You then have two parties (transferor and transferee) claiming that a transfer has taken place, and one party (RIPE, etc.) claiming that it has not. Instead of a clear statement of the legitimate user, you need to go to some kind of adjudication panel to sort the mess out.
2) Transparency
I fully agree on transparency. What I fail to see is what is meant with 'public way' and why 2007-08 would fall short of that mark.
Because under 2007-08 two LIRs can secretly negotiate and execute a transfer or addresses under some secret set of terms which may or may not include payments of cash. RIPE only gets involved at the end of the process to update the database, and the terms of the exchange are never disclosed.
When IPv4 space moves hands because of a merger or acquisition this doesn't happen either. That's all current bottom-up policy.
During M&A activity, it is public who receives the functioning network assets, and that is the criteria (technical justfication) for allocating IP addresses to the new network owner. The IP addressing aspects are still open and public.
3) Fair and neutral reuse All recovered address space should be recycled in a globally synchronized fashion. RIRs should work harder recovering IPv4 space. I think that this is a good idea, but let's not overestimate what IANA and the RIRs are able to recover.
Under 2007-08, you are right that the RIRs will not recover much. But if we leave the transfer framework as it currently is, then most IPv4 addresses that become surplus due to IPv6 deployment, will go back to the RIRs. Here ETNO is supporting the current framework because it provides fair and neutral terms for reuse of addresses. To get addresses, everyone must provide a technical justification. When that justification goes away, everyone must return the addresses to RIPE unless there is some imminent new technical justification in their own organization, for instance shutting down dialup services in a company who is still growing VPN or Internet access or broadband business. Reuse, happens naturally and no new policies are needed. It would be nice for RIPE to remind LIRs about returning IP addresses and set some guidelines such as don't bother until you have more than a /24, if you expect an adjacent block to become free within 6 months then wait before returning, etc. That could be done without making fundamental changes to the transfer framework.
I find it curious that ETNO puts significant effort into writing position papers such as this one and not join us in that bottom-up process, but that's a different matter.
ETNO is just a bunch of people, much like RIPE. Different people have different working styles and that's OK. Note that more people from ETNO companies are starting to get involved in RIPE as they learn and understand RIPE's working practices. May I suggest that RIPE could help by offering to present a session at ETNO's next annual conference in November.
5) No connection between IPv4 and IPv6 Policies we make for the 'afterlife' of IPv4 should not affect any policy with regard to IPv6. Yes. 2007-08 is in compliance with that.
Unfortunately, 2007-08 will lead many to think that selling address blocks is OK in the RIPE region. As a result, this does affect IPv6 since if an IPv4 address block is a saleable asset then IPv6 addresses are also saleable assets.
In describing transfers, ETNO argues that transfers are the first step towards a market. That is false. It is the first step towards a market under the control of the RIRs and the current bottom-up policy process we all know and love. A market will evolve whether we accommodate it or not.
Focussing just on the last sentence it is clear that you believe a market will evolve, and that 2007-08 "accomodates" a market. That is precisely where ETNO disagrees. ETNO believes that a market can be prevented from evolving if RIPE takes the correct actions, and that "accomodating" a market is a wrong action for RIPE to take. Remember that collectively, ETNO members have a huge number of IPv4 addresses allocated. If they do not participate in any attempts at forming a market, then the market will be stunted, and lack liquidity. In addition, when the shift to IPv6 begins in earnest, because vendors have fixed the problems which are still blocking it, ETNO member companies will be deploying IPV6 rather quickly. This will result in a significant percentage of their IPv4 addresses returning to RIPE for reallocation. That event will kill any existing market for IPv4 addresses. So the best possible scenario for an IPv4 market will be a trickle of transactions put on for show then a runout of IANA's free pool followed by a sharp increase in small block transactions by some players trying out the market. At this point prices will sharply increase causing buyers to back out of the market unless they are in desperate straits. Then IPv6 deployers will begin giving some addresses back to RIPE at which point the market collapses. Frankly, it would be better for this to play itself out in a black market scenario without RIPE involvement.
Inaction is not a sign of good stewardship. So we need to be seen doing something, preferably something that will make a difference.
You are right. It is time for RIPE to take more action to help LIRs audit their supply of addresses and to set some reasonable guidelines for returning unused IPv4 addresses. Since addresses are only usable in aggregate, and since LIRs could potentially increase those aggregate block sizes with sensible internal processes, it is time for RIPE to work on a best practices document for IPv4 addressing.
I would urge ETNO to become part of this discussion and work with the community in our prized bottom-up policy process, rather than setting limits to what can and can not be discussed an not coming up with solutions themselves.
Note that Mark McFadden, who presented an ETNO document a couple of RIPE meetings ago, and myself, both work for an ETNO member. I agree that it would be nicer to see more ETNO members join in the discussion, but please recognize that ETNO deals with far more issues and policy organizations than just RIPE. I suspect that many ETNO members see RIPE as a functioning organization that is mostly doing good work for the network, and therefore RIPE doesn't need as much attention as some other organizations that make laws and regulations (which have almost the same force as a law). --Michael Dillon
Hi Michael, Just browsing through your response one item immediately sticks out: is there a commitment from ETNO in here that their members would immediately return unused IPv4 space ? That'd be great - if IPv6 were to be deployed fully by all ETNO members in the next 2 years, which you think is easily achievable, we'd never run out of space. I'm asking because it is exactly the opposite of what representatives of a lot of ETNO members have been telling me in private. Kind regards, Remco van Mook (again no hats)
-----Original Message----- From: address-policy-wg-admin@ripe.net [mailto:address-policy-wg-admin@ripe.net] On Behalf Of michael.dillon@bt.com Sent: woensdag 4 juni 2008 12:59 To: address-policy-wg@ripe.net Subject: RE: [address-policy-wg] 2007-08: response to ETNO position
Let me first rephrase what 2007-08 is about, before responding to ETNO's position. 2007-08 is about setting up a framework for (IPv4) resource transfers between LIRs. Nothing more, nothing less. No IPv6, no ASNs. No PI, no ERX. It is a framework in the sense that, in my opinion, it encompasses the absolute minimum set of policy changes required for this to happen without immediately shooting ourselves in the foot. It is a starting point, not an end point for transfer policies.
However, we already have a framework for IP address transfers between LIRs which works substantially the same in all RIR regions. If it involves a change in corporate ownership, you simply need to show proof of the change and that the network is still operating under the new entity. The RIR then updates all records showing the new ownership. Or you give the address block back to the RIR because you no longer need it, and some other LIR will get those addresses at some point in the future.
This framework works. There is no reason to believe that it would cease working as IPv6 deployment picks up. Most network operators are undertaking some form of audit of their internal use of IP addresses in preparation for the shortage, which means that as IPv6 picks up, they should know when they have surplus IPv4 addresses and can return them to RIPE or whatever RIR they came from. If it is not broken, then why would we want to change this framework?
You claim that this will speed up the process of transfers but I cannot see how adding a new set of transfer rules will speed up anything. Even though you claim that this is an LIR-to-LIR transfer without RIPE involvement, this is not true since there are various rules which RIPE must administer and RIPE can disallow a transfer if, for instance, an LIR has received a re-allocation within the past 24 months.
1) Identification of legitimate use
While I agree that we need to keep track on who has rights on what - it is one of the key points of 2007-08 - doing it globally would mean abandoning all RIR databases for a global registry.
Huh? I can't see where you get this idea. This ETNO point is in full support of the existing RIR system, where every IP address range is under the authority of one of the RIRs, and where the identity of the "legitimate user" can be authoritatively discovered in the RIR's database (or whois directory).
In this point ETNO is supporting the existing transfer framework in which all IP address transfers must go through one of the 5 RIRs. Under 2007-08 it is possible that some LIRs will transfer addresses which an RIR will not register in the database. You then have two parties (transferor and transferee) claiming that a transfer has taken place, and one party (RIPE, etc.) claiming that it has not. Instead of a clear statement of the legitimate user, you need to go to some kind of adjudication panel to sort the mess out.
2) Transparency
I fully agree on transparency. What I fail to see is what is meant with 'public way' and why 2007-08 would fall short of that mark.
Because under 2007-08 two LIRs can secretly negotiate and execute a transfer or addresses under some secret set of terms which may or may not include payments of cash. RIPE only gets involved at the end of the process to update the database, and the terms of the exchange are never disclosed.
When IPv4 space moves hands because of a merger or acquisition this doesn't happen either. That's all current bottom-up policy.
During M&A activity, it is public who receives the functioning network assets, and that is the criteria (technical justfication) for allocating IP addresses to the new network owner. The IP addressing aspects are still open and public.
3) Fair and neutral reuse All recovered address space should be recycled in a globally synchronized fashion. RIRs should work harder recovering IPv4 space. I think that this is a good idea, but let's not overestimate what IANA and the RIRs are able to recover.
Under 2007-08, you are right that the RIRs will not recover much. But if we leave the transfer framework as it currently is, then most IPv4 addresses that become surplus due to IPv6 deployment, will go back to the RIRs. Here ETNO is supporting the current framework because it provides fair and neutral terms for reuse of addresses.
To get addresses, everyone must provide a technical justification. When that justification goes away, everyone must return the addresses to RIPE unless there is some imminent new technical justification in their own organization, for instance shutting down dialup services in a company who is still growing VPN or Internet access or broadband business. Reuse, happens naturally and no new policies are needed.
It would be nice for RIPE to remind LIRs about returning IP addresses and set some guidelines such as don't bother until you have more than a /24, if you expect an adjacent block to become free within 6 months then wait before returning, etc. That could be done without making fundamental changes to the transfer framework.
I find it curious that ETNO puts significant effort into writing position papers such as this one and not join us in that bottom-up process, but that's a different matter.
ETNO is just a bunch of people, much like RIPE. Different people have different working styles and that's OK. Note that more people from ETNO companies are starting to get involved in RIPE as they learn and understand RIPE's working practices. May I suggest that RIPE could help by offering to present a session at ETNO's next annual conference in November.
5) No connection between IPv4 and IPv6 Policies we make for the 'afterlife' of IPv4 should not affect any policy with regard to IPv6. Yes. 2007-08 is in compliance with that.
Unfortunately, 2007-08 will lead many to think that selling address blocks is OK in the RIPE region. As a result, this does affect IPv6 since if an IPv4 address block is a saleable asset then IPv6 addresses are also saleable assets.
In describing transfers, ETNO argues that transfers are the first step towards a market. That is false. It is the first step towards a market under the control of the RIRs and the current bottom-up policy process we all know and love. A market will evolve whether we accommodate it or not.
Focussing just on the last sentence it is clear that you believe a market will evolve, and that 2007-08 "accomodates" a market. That is precisely where ETNO disagrees. ETNO believes that a market can be prevented from evolving if RIPE takes the correct actions, and that "accomodating" a market is a wrong action for RIPE to take.
Remember that collectively, ETNO members have a huge number of IPv4 addresses allocated. If they do not participate in any attempts at forming a market, then the market will be stunted, and lack liquidity. In addition, when the shift to IPv6 begins in earnest, because vendors have fixed the problems which are still blocking it, ETNO member companies will be deploying IPV6 rather quickly. This will result in a significant percentage of their IPv4 addresses returning to RIPE for reallocation. That event will kill any existing market for IPv4 addresses.
So the best possible scenario for an IPv4 market will be a trickle of transactions put on for show then a runout of IANA's free pool followed by a sharp increase in small block transactions by some players trying out the market. At this point prices will sharply increase causing buyers to back out of the market unless they are in desperate straits. Then IPv6 deployers will begin giving some addresses back to RIPE at which point the market collapses.
Frankly, it would be better for this to play itself out in a black market scenario without RIPE involvement.
Inaction is not a sign of good stewardship. So we need to be seen doing something, preferably something that will make a difference.
You are right. It is time for RIPE to take more action to help LIRs audit their supply of addresses and to set some reasonable guidelines for returning unused IPv4 addresses. Since addresses are only usable in aggregate, and since LIRs could potentially increase those aggregate block sizes with sensible internal processes, it is time for RIPE to work on a best practices document for IPv4 addressing.
I would urge ETNO to become part of this discussion and work with the community in our prized bottom-up policy process, rather than setting limits to what can and can not be discussed an not coming up with solutions themselves.
Note that Mark McFadden, who presented an ETNO document a couple of RIPE meetings ago, and myself, both work for an ETNO member. I agree that it would be nicer to see more ETNO members join in the discussion, but please recognize that ETNO deals with far more issues and policy organizations than just RIPE. I suspect that many ETNO members see RIPE as a functioning organization that is mostly doing good work for the network, and therefore RIPE doesn't need as much attention as some other organizations that make laws and regulations (which have almost the same force as a law).
--Michael Dillon
Any opinions expressed in the email are those of the individual and not necessarily of the company. This email and any files transmitted with it are confidential and solely for the use of the intended recipient and do not constitute an offer or acceptance by Equinix, Inc., Equinix Europe Ltd or any of their group entities to buy or sell any products or services in any jurisdiction. If you have received this email in error please delete this email immediately and notify the IT manager.
This communication is sent on behalf of one of the European entities in the Equinix, Inc. Group. The ultimate holding company in Europe is Equinix Europe Ltd whose registered address is Quadrant House, Floor 6, 17 Thomas More Street, Thomas More Square, London E1W 1YW and the Company's registered number is 6293383. The registration details of other Group entities are available at www.eu.equinix.com
Just browsing through your response one item immediately sticks out: is there a commitment from ETNO in here that their members would immediately return unused IPv4 space ?
ETNO can't make that kind of commitment. If RIPE wanted to ask its LIRs to make such a commitment, then that would be interesting. However, my comment is an attempt to clarify that the current framework for IP address block transfers does include returning blocks to RIPE for reallocation.
That'd be great - if IPv6 were to be deployed fully by all ETNO members in the next 2 years, which you think is easily achievable, we'd never run out of space.
I never said anything about full deployment of IPv6. All we need in order for IPv4 address blocks to be returned is for IPv6 deployment to be slightly more than normal network growth. If you assume that IPv4 deployment has been going on for about 10 years, then normal growth would not be more than one tenth of full deployment. Also note that even less IPv6 deployment than that, will be enough to reduce the consumption of IPv4 address blocks and thus, extend the lifetime of IPv4 well beyone 2 years. In fact, we may never run out of IPv4 addresses if IPv6 deployment happens fast enough. Since new product development status is generally considered highly confidential, we really don't know how much any companies have done to prepare for this so we really cannot reasonably estimate how long it will take. I'm not sure if anyone has paid the research companies like Gartner, Forrester, Ovum, etc to go around and check on the status of IPv6. It would be interesting to know if anyone has any fact-based estimates of how long it would take the IP network industry to deploy enough IPv6 service to cut their IPv4 consumption in half, and therefore push the end-date for IPv4 out in the future, 5 to 6 years from now.
I'm asking because it is exactly the opposite of what representatives of a lot of ETNO members have been telling me in private.
I'm not sure that most of the ETNO representatives would be in a position to know this within their companies. And if they did know, I expect most of them would not be allowed to say anything definite. And of course, if you ask the wrong questions, then any answers are worthless. I don't think that any major ISP in any country would be able (technically and commercially) to deploy IPv6 FULLY before 5 years from now. That's based on the history of IPv4 deployment and how long it took for various networking issues to be solved to the point where most people receive a fast and reliable Internet access service. Or IP-VPN service. Or private IP internetwork service. But we don't need full deployment in order to address the problem of IPv4 exhaustion. We just need enough deployment to slow down IPv4 consumption, and we need that deployment to be continuous with a slightly increasing rate so that it becomes business-as-usual for most IP network operators. There will be a period where vendors of routers, firewalls, access gateways, and software work through a long list of minor (and occasionally major issues) which block deployment in one way or another. That process is what will take about 5 years to get through at which point some network operators (not all of them) will be able to fully deply IPv6 to all of their PoPs and across all of their acess technologies. It doesn't help anything when people demand that network operators accomplish the impossible. This is a shared journey that requires cooperative work from a lot of organizations in order to make IPv6 fully usable. There is no doubt that we will all take that journey, but not everybody travels at the same speed. --Michael Dillon P.S. here's a quote from a message that I posted to ARIN's policy mailing list about a year ago. I expect that we will see 3-4 years of IPv6 deployment, followed by 10 years or so in which IPv4 usage steadily reduces and IPv6 steadily rises. Then in about 14-15 years we will begin a period of consolidation in which there is an effort to retire IPv4 on the public Internet and in corporate/enterprise networks. Give this another 10 years, which means that in about 25 years, IPv4 will be considered end-of-life. At that point, general purpose networking gear will cease to support IPv4, however it is possible that new IPv4 devices will continue to be built for special uses. Even today, it is not hard for an undergraduate student to build their own IPv4 router using off-the-shelf parts and open-source software. Even ASICs are available off-the-shelf in the form of FPGAs and one can expect that in 25 years this type of thing becomes even easier and cheaper. So IPv4 may linger on for a few generations hidden inside MP3 player headsets and heel computers (inside your running shoes) and concert posters.
On 4 jun 2008, at 12:59, <michael.dillon@bt.com> <michael.dillon@bt.com> wrote:
Let me first rephrase what 2007-08 is about, before responding to ETNO's position. 2007-08 is about setting up a framework for (IPv4) resource transfers between LIRs. Nothing more, nothing less. No IPv6, no ASNs. No PI, no ERX. It is a framework in the sense that, in my opinion, it encompasses the absolute minimum set of policy changes required for this to happen without immediately shooting ourselves in the foot. It is a starting point, not an end point for transfer policies.
However, we already have a framework for IP address transfers between LIRs which works substantially the same in all RIR regions. If it involves a change in corporate ownership, you simply need to show proof of the change and that the network is still operating under the new entity. The RIR then updates all records showing the new ownership. Or you give the address block back to the RIR because you no longer need it, and some other LIR will get those addresses at some point in the future.
This is hardly the same, your talking about mergers in which one of the LIR's will be closed at the end of the procedure. Further this procedure (ripe-301) doesn't talk about reclaiming address, to quote the docuent: "Transferred allocations containing a large amount of unassigned address space may be set aside and kept by the RIPE NCC until the other allocations held by the LIR are considered fully used (about 80%). Once the LIR has reached full utilisation in its other allocations, the reserved allocations will be made available to the LIR." Which, if you're building reserves for after 2012 is acceptable and might even help getting away from the tax collectors, should there ever be such time (which I hope will never happen) where IPv4 addresses are considered a taxable asset.
This framework works. There is no reason to believe that it would cease working as IPv6 deployment picks up. Most network operators are undertaking some form of audit of their internal use of IP addresses in preparation for the shortage, which means that as IPv6 picks up, they should know when they have surplus IPv4 addresses and can return them to RIPE or whatever RIR they came from. If it is not broken, then why would we want to change this framework?
Wether we like it or not, there will eventually be some form of market. It might even be your scenario where people simply buy the whole company, including the registry, to get their hands on unused resources or where the value of the company will be depending on the number of unused and available resources. Should this situation become reality, I don't see much people giving back stuff which they can sell for cold hard cash.
You claim that this will speed up the process of transfers but I cannot see how adding a new set of transfer rules will speed up anything. Even though you claim that this is an LIR-to-LIR transfer without RIPE involvement, this is not true since there are various rules which RIPE must administer and RIPE can disallow a transfer if, for instance, an LIR has received a re-allocation within the past 24 months.
Which will protectr the market, should there be any, from speculation and maybe even from emerging at all.
2) Transparency
I fully agree on transparency. What I fail to see is what is meant with 'public way' and why 2007-08 would fall short of that mark.
Because under 2007-08 two LIRs can secretly negotiate and execute a transfer or addresses under some secret set of terms which may or may not include payments of cash. RIPE only gets involved at the end of the process to update the database, and the terms of the exchange are never disclosed.
And where exactly does this differ from ripe-301 ?
When IPv4 space moves hands because of a merger or acquisition this doesn't happen either. That's all current bottom-up policy.
During M&A activity, it is public who receives the functioning network assets, and that is the criteria (technical justfication) for allocating IP addresses to the new network owner. The IP addressing aspects are still open and public.
I find this a pretty big claim, M&A procedures aren't that public, if your lucky both parties involved have to comply to stock market regulations, but still it usually is only announced when the deal is already made. I also can't see, not being involved directly, which of those transferred resources are being put aside by the NCC, the only thing I can make up from public records is there has been some form of merger and resources are now assigned to a new LIR and maybe I can find a newspaper article on what happened to the company. I also can at least give you one example where addresses have been transferred between two LIR's in the aftermath of a bankrupcy, where 301 wasn't applied and both the losing as the recieving registry kept operating indepently.
3) Fair and neutral reuse
All recovered address space should be recycled in a globally synchronized fashion. RIRs should work harder recovering IPv4 space.
I think that this is a good idea, but let's not overestimate what IANA and the RIRs are able to recover.
Under 2007-08, you are right that the RIRs will not recover much. But if we leave the transfer framework as it currently is, then most IPv4 addresses that become surplus due to IPv6 deployment, will go back to the RIRs. Here ETNO is supporting the current framework because it provides fair and neutral terms for reuse of addresses.
To get addresses, everyone must provide a technical justification. When that justification goes away, everyone must return the addresses to RIPE unless there is some imminent new technical justification in their own organization, for instance shutting down dialup services in a company who is still growing VPN or Internet access or broadband business. Reuse, happens naturally and no new policies are needed.
It would be nice for RIPE to remind LIRs about returning IP addresses and set some guidelines such as don't bother until you have more than a /24, if you expect an adjacent block to become free within 6 months then wait before returning, etc. That could be done without making fundamental changes to the transfer framework.
I think this is a prime example on where theory and everyday life are completely different, there isn't much space returned and knowing that once you return resources it will only get harder to ever get them back will only hold people back from returning address space.
5) No connection between IPv4 and IPv6
Policies we make for the 'afterlife' of IPv4 should not affect any policy with regard to IPv6.
Yes. 2007-08 is in compliance with that.
Unfortunately, 2007-08 will lead many to think that selling address blocks is OK in the RIPE region. As a result, this does affect IPv6 since if an IPv4 address block is a saleable asset then IPv6 addresses are also saleable assets.
Let's, for the sake of the argument, assume IPv6 is indeed infinite (ignoring the 'classfull' behaviour in current v6 policies), why would I ever want to buy IPv6 addresses as I can get brandnew ones for almost free at the RIR around the corner ? And if there is no reason to buy, there won't be a reason to sell either.
In describing transfers, ETNO argues that transfers are the first step towards a market. That is false. It is the first step towards a market under the control of the RIRs and the current bottom-up policy process we all know and love. A market will evolve whether we accommodate it or not.
Focussing just on the last sentence it is clear that you believe a market will evolve, and that 2007-08 "accomodates" a market. That is precisely where ETNO disagrees. ETNO believes th000at a market can be prevented from evolving if RIPE takes the correct actions, and that "accomodating" a market is a wrong action for RIPE to take.
Remember that collectively, ETNO members have a huge number of IPv4 addresses allocated. If they do not participate in any attempts at forming a market, then the market will be stunted, and lack liquidity. In addition, when the shift to IPv6 begins in earnest, because vendors have fixed the problems which are still blocking it, ETNO member companies will be deploying IPV6 rather quickly. This will result in a significant percentage of their IPv4 addresses returning to RIPE for reallocation. That event will kill any existing market for IPv4 addresses.
So the best possible scenario for an IPv4 market will be a trickle of transactions put on for show then a runout of IANA's free pool followed by a sharp increase in small block transactions by some players trying out the market. At this point prices will sharply increase causing buyers to back out of the market unless they are in desperate straits. Then IPv6 deployers will begin giving some addresses back to RIPE at which point the market collapses.
Frankly, it would be better for this to play itself out in a black market scenario without RIPE involvement.
And in the mean time also the ETNO members get spammed/scanned/hacked from one of those black market address blocks. Who do you turn to ? The original LIR ? What if it's a part of a pre-RIR block ?
Inaction is not a sign of good stewardship. So we need to be seen doing something, preferably something that will make a difference.
You are right. It is time for RIPE to take more action to help LIRs audit their supply of addresses and to set some reasonable guidelines for returning unused IPv4 addresses. Since addresses are only usable in aggregate, and since LIRs could potentially increase those aggregate block sizes with sensible internal processes, it is time for RIPE to work on a best practices document for IPv4 addressing.
It's probably unnecessary to state I'm also against the active development of a market, I do however support 2007-08 because reality tells me there already is a market, people are requesting PI spae for a living. I would rather see the RIR's as a title agent so that market is as transparent as it can be and we can even put a dampening effect on it, by for instance prohibiting the re-transfer of a block wihtin 24 months. Finally on your comment on auditing address space needs, doing a little digging on the database for one of the etno members, I would love to see that /11 return to the unallocated pool. Groet, MarcoH (no hats, just hair)
participants (4)
-
Marco Hogewoning -
michael.dillon＠bt.com -
Remco van Mook -
Remco van Mook (Virtu)