Draft transcript

Policy SIG - Session One

Thursday 6 September 2007

1130-1300

TOSHIYUKI HOSAKA:

Hello, everyone. I'm Toshiyuki Hosaka. Policy SIG is likely to begin in a few moments. So I appreciate you could have a seat and come back from the coffee break.

I think we are prepared to start the Policy SIG session. Hello, everyone once again. I'm Toshiyuki Hosaka, APNIC Policy SIG chair. Let's start the Policy SIG.

This is the agenda today and at the top of the agenda we have a committee resolution proposal. This is from Randy Bush, IIJ. Current IPv4 and IPv6 issues. I will call on Randy to come up here to introduce this resolution to the community.

Proposed APNIC community resolution on current IPv4 and IPv6 issues

RANDY BUSH:

Hi. I'm Randy Bush from IIJ. Paul Wilson actually, according to the drafting of this resolution. But I am informally acting as the - you can read it there or you can read it on the website if you're a little far back. But it essentially says that we recognise the run-out of the IANA v4 free pool and we realise that IPv6 is critical, in a critical transition. And that during this period I'm going to read directly from it, "We will be learning and adapting and that address management policies may also change to adapt to due circumstances." And we reassert our support for open, bottom-up consensus-based decision-making but we also call on the leading senior expert members of this community to provide leadership for solutions in this space. So, this is essentially a statement of philosophy. And it is. We're running out of the free pool of v4, life is going to change. We would like to transition to v6. We're going to do it with the community in consensus and hopefully we have technical and policy guidance from the senior people in this community.

So this kind of sets the mood and the framework, which is why I asked to have this proposal put first, because it kind of sets the philosophy we're going to go forward over the next difficult years. Thank you.

TOSHIYUKI HOSAKA:

Thank you, Randy. I think this is a very important resolution coming from our community. Let me repeat this draft resolution. Let me read this resolution. The APNIC community resolves as follows. We recognise that at current rates of allocation, the remaining free pool of IPv4 address space will be consumed within the next 2-4 years.

We agree that this situation requires a concerted effort by this community, working for the common good, to seek, examine and adopt responsible measures for the management of remaining IPv4 address space. We recognise that during this period, we will be learning and adapting, and that address management policies may also change to adapt to new circumstances.

We recognise the critical importance of IPv6 to the future success of the Internet and will actively promote the adoption of IPv6 and focus our efforts towards comprehensive deployment of IPv6 in the Asia Pacific region.

We reassert our support for open, bottom-up and consensus-based decision-making but we also call upon the leading senior and expert members of this community to provide strong leadership in the search for solutions to these issues of IPv4 address management and transition to IPv6, both within the Asia Pacific region and globally.

So this is not from our policy proposal and if we do not have substantial or strong objection to this, we would like to adopt, resolve this resolution as written. Is there any objections? I think no. So it is resolved as written. Thank you very much. APPLAUSE

Policy SIG Co-chair election

So next item on the agenda is the Policy SIG co-chair election. We have two seats of the co-chair of Policy SIG and have only two nominations from communities. The nomination is from Randy Bush and Jan Zheng. If you don't have any objections to those two people, I would like to welcome them to the co-chair. Is there anybody who has objections to those people? I think none. So please welcome to new co-chairs, Randy Bush and Jan Zheng. Would you like to say a few words as a new co-chair?

RANDY BUSH:

Haven't people already heard enough from me? I'm an Internet researcher and operator. I've been doing this stuff for 25 years. I've been doing computing for over 40. An old dog trying to learn new tricks.

TOSHIYUKI HOSAKA:

Jan.

JAN ZHENG:

Hi. I'm a new face here. I'll be working on domestic and international policy in CNNIC. Before I joined CNNIC, I was working for Cisco Systems. I am always fascinated by the new technologies, Internet, which profoundly changed people's life. So I really want to make more contribution to our APNIC community. So I would like to serve on this open policy SIG. Thank you.

TOSHIYUKI HOSAKA:

Welcome to the new co-chairs. Thank you.

So the next agenda is IPv6 technical SIG chair election. Since we have no IPv6 technical SIG in this APNIC 24 meeting, we are giving 10 minutes or so for the IPv6 SIG chair election. I will hand over now to Kazu Yamamoto.

IPv6 technical SIG Chair election

KAZU YAMAMOTO:

Good morning, everybody. As I explained six months ago, we IPv6 technical SIG started with co-chair and SIG chair election. We select chair of IPv6 technical SIG. I have been working as IPv6 technical SIG chair for five years, if I remember correctly. That is a little bit long so it's a very good time for me to step down. Fortunately we have two nominations at this time. Let me call to nominations. The first one, Ching-Heng Ku, please come up. And the second person is Yoshinobu Matsuzaki. I don't worry about which person is elected. But because the position is one, we have to choose one person for the chair. So if you're not elected this time, I would encourage you to come back one year later to be nominated in the SIG as co-chair. OK.

CHING-HENG KU:

Good morning. I have some slides to introduce myself. I'm the director of the IP Department of TWNIC. My IPv6 background includes currently I am the co-chair of Taiwan IPv6 development and deployment program. I am in charge of the project leader of the IPv6 tunnel broker services for ISPs in Taiwan. Besides, I am a secretary on the Secretariat of the Asia Pacific IPv6 task force for two years and I coordinated the Asia Pacific IPv6 summit and coordinated the Asia Pacific task force. My background include I am the chair of the NIR host master, workshop and keep involving in APNIC meetings for four years. I host the APNIC 22 Kaohsiung meeting in 2006. My motivation for IPv6 technical SIG chair is to enhance the awareness and the importance of IPv6 and to promote IPv6 technology and allocations in the Asia Pacific regions and how to connect and coordinate the cooperation between AP - Asia Pacific regions and other continents and to assist the transition from IPv4 to IPv6 smoothly. And these are my brief introduction. Thank you very much.

YOSHINOBU MATSUZAKI:

I'm a senior engineer at IIJ. Specialising in routing, DNS and security. I operate its backbone network for 10 years. Maintained IIJ's IPv6 backbone since 1999. I have been an IPv6 home user since 1999. I have given numerous talks at JANOG, NANOG, RIPE, APNIC and other meetings. And I have attended at the APNIC since APNIC 19. And at this moment, I present my analysis about the DNS server. I'm a member of the Japanese study group of IPv6 transition. And I have got approval from my boss to act as the chair of IPv6 technical SIG so I can come to further APNIC meetings as well and I believe I can contribute to IPv6 technical SIG and, again, this is Yoshinobu Matsuzaki. Thank you.

KAZU YAMAMOTO:

So please raise your hand when I call the name for who you support. OK.

So please raise your hand if you support Ching-Heng Ku? Please raise your hand. Higher. 11. So thank you for your support and we'll conclude, OK. So please raise your hand if you support Mr Matsuzaki? So 29 people support Mr Matsuzaki. So a big applause, please. APPLAUSE

So I step down at this time. As I said before, I have been working as chair for five years and I like APNIC very much because many people are kind and I had a very good experience. Thank you very much. Mr Matsuzaki, we'll give some words to you.

YOSHINOBU MATSUZAKI:

Thank you. I have already introduced myself, so I'll do my best for IPv6 technical SIG. Thank you.

KAZU YAMAMOTO:

Thank you very much.

Open Action Items

TOSHIYUKI HOSAKA:

Thank you very much. Let's go back to the agenda. So I want to check the open agenda action items for the policy SIG. You can see the policy documentation on those URLs. And your involvement is indispensable. For this policy development process. Housekeeping - speakers, please speak slowly and clearly. Many participants, English is not their first language. The session is being web cast. Remote participation is possible. Participate openly. Speakers from the floor must use the mic to speak. Decisions by consensus, not a vote. Agreement and on compromise.

We are checking open action items. Policy 22-005 pending approval at each remaining stage of the policy proposal process, APNIC Secretariat to implement the proposal prop-035, IPv6 portable assignment for multihoming. This was implemented in March. It's done.

Policy 22-006 - pending approval at each remaining stage of the policy proposal process, APNIC Secretariat to implement the proposal prop-033, end-site assignment policy for IPv6. This was implemented in March this year so it's done.

The next is policy 23-001 - chair to refer prop-042 - proposal to change IPv6 initial allocation criteria to mailing list for further discussion. It's done. I have sent that to this proposal to the mailing list for final discussion. And this proposal is on the agenda today.

The policy 23-002, chair to refer proposal prop-043, proposal to remove reference to IPv6 policy document as an interim policy document to mailing list for further discussion. This is done.

Policy 23-003 - chair to refer prop-046 - IPv4 countdown policy proposal to mailing list for further discussion. It's on the agenda today. Policy 23-004 - chair to refer proposal-037 - deprecation of email updates for APNIC registry and Whois data to mailing list for further discussion. This proposal is not on the agenda today. Policy 23-005 - Marshall Eubanks to present prop-047 - eGLOP multicast address assignments as a formal policy proposal at APNIC 24. This remains open. Any comments or questions? No. OK. Thank you very much. So I will move to the next agenda. I have plan to request for Sunny to present global policy updates on the next agenda. I will skip this one to the next session and I will call on Izumi Okutani from JPNIC to present the proposal 046, IPv4 countdown policy.

I'm one of the co-authors of this proposal, so I will get Randy to chair this particular proposal and the next one from Haitham.

Prop-046: IPv4 countdown policy

IZUMI OKUTANI:

Good morning, everyone. My name is Izumi Okutani and I'm from JPNIC which is the National Internet Registry in Japan. In my presentation today I would like to discuss what would be the possible issues we would be facing at the time of IPv4 address exhaustion and if there's anything we can do in terms of policy to help make distribution smooth at the time of the exhaustion? And my presentation consists of two parts. Firstly, it's informational presentation to discuss about what we can do to distribute the last piece or the last pieces of APNIC address pool. And the next part will be global policy proposal discussing about the distribution of IANA pool to RIRs.

So start with the first part about the distribution of APNIC pool. Before going in to the details, I would like to share the background of the situation surrounding the exhaustion. I'm sure many of you is already aware of the situation. This is a quote from potaroo.net. It's expected that currently we have 46 /8s available in IANA in an allocated pool. And based on the consumptions at the moment, it is expected that by May of 2010, the IANA pool will run out. So there will be no more free pool available in IANA to be distributed to RIRs, which is not very long from now. It's only less than three years from now. So what we've done in Japan until this point is we've set up an expert team to discuss what would be the possible issues we would be facing. And if there's anything we can do in pall it's to help reduce or address the issues that would come up at the time of exhaustion.

And we've also been holding open policy meetings which we have within Japan similar to what we have here but it's like a smaller JP version. And the last open policy meeting we had discussions based on a draft proposal on development by this expert team. And the general feeling there was that we did have consensus that we would probably need to have changed the current policy to repair for the exhaustion and there was also general support to restrict the current distribution criteria after a certain date. However, it's quite expectedly and naturally, there were so many different ideas on how we should restrict the distribution, so some people felt that we should give priority to newcomers, while the others felt we should stick with the existing policy, etc, etc. So what I'd like to do in this presentation is to share what were the options that were mentioned in our meeting, as well as some of the discussions in other RIRs' mailing lists. And consider what would be the best possible solution or action we could take in this region.

So to put it in short, this presentation intends to review what would be the most effective policy measure in APNIC region to prepare for IPv4 exhaustion which is expected to come up in a few years. It should be read in conjunction with a policy proposal of 046 which I will present after this presentation. So it's like a whole set.

So what are the options? We came up with five listed here. And the first is to do nothing about it with an explicit decision and continue distribution under the current scheme. And the second option would be to restrict the criteria, so the people who will be receiving the v4 address space will be restricted in some ways to promote efficient utilisation of the remaining pool. And the third option would be - don't worry too much about efficient utilisation because v4 will run out at some point or the other, so let's focus on the next step which would be v6 deployment. And there should be something we should do to promote this deployment from policy area as well. And the remaining two options - D and E - they're not related but trying to address the same issue, which is once v4 address runs out, those without a single IPv4 global address space, they have no means to connect to the v4 Internet, which is currently the Internet at the moment. So some people feel that we should do something about it to help those people. And option D is trying to tackle this issue very directly by giving priority to initial allocations - i.e., those without the global v4 space at the moment. And the second option is rather than giving priorities to newcomers, they think, "OK, if there is a translator network, provides smooth translation between v4 and v6, then individual networks don't have to worry about IP networks. So those without a single v4 space can still connect to v4 networks, so we should give priority to those translators.

Let me just explain the pros and cons of the options. Do nothing about it as an explicit decision. It would be a problem if we don't realise the situation and do nothing and, you know, at the time of exhaustion think, "Oh, facing the problem." That would be a problem. After exploring different possible options and then balancing the pros and cons of these options and explicitly feel it is indeed the best to keep up with the current criteria and continue with the current policy. That is certainly one wise decision to make. And the advantage of this would be that it's simple and easy. People won't know what to expect. People need to think, "OK, the criteria will remain the same. Remain unchanged until, you know, the last piece of v4 address space is available within RIRs. And another thing is that if we change the criteria, there will always be people who are left out from receiving the distribution. So those people will always feel that it is this new change policy will be unfair and it is really difficult to, you know, it gets in to discussions of who we should give the priority to, it can go in parallel and sometimes be controversial. So we can see this feeling of unfairness by maintaining the current criteria. However, the possible downside could be that, as one of the possible issues I mentioned - those without a single global v4 address space have no means to connect to the v4 Internet. And maybe this could be solved operationally but it could take a few years, maybe more than at the time of exhaustion. And then it might be we might have no means to help those people. And if we can help much more easily by implementing a policy for those without global address space, is it really wise to opt out this option? Maybe we should be open to other possible options that's available in policy area. So that would be a possible downside of this option.

And another group of people they feel that since v4 address space will be scarce, so we should promote more efficient utilisation of the remaining space by tightening the criteria for distribution. For example, raising the utilisation rate which is currently 80% for subsequent allocation but raising to 95%, 100%. Or requiring more documentation, we have to give invoices for each of the requests or whatever.

And the good thing about it, it would probably help encourage much efficient utilisation of v4 and might prolong the life of v4. But what we should be careful about is it would be a lot of burden on ISPs as well as host masters to evaluate the criteria and that would be much more strict and would it be worth going through the efforts if this is maybe a single /8 globally which can only prolong the life for maybe less than a month or so. So we have to balance out the efforts and the results we will get as a result of this. And there could be other possible options that would encourage efficient utilisation of address space as maybe recovery of historical address space or what this could be controversial, but allowing transfers between the existing v4 orders etc, etc. These are the possible downsides.

The other idea is to promote v4 aspect as well. So those who will be requesting v4 will also need to be able to justify that they plan to provide commercial v6 service within one year, for example. So implementing a policy like this to promote the deployment of v6. And I suppose the advantage of this option would be rather than leaving it up to ISPs to make the decision, it allows them to get prepared for the v6 which they might seem to move in to it in a few years anyway. So much more in advance than at the time of exhaustion. But then the possible counterargument against this could be that it's far better to leave it up to the decision of individual ISPs, by artificially trying to promote the deployment of v6. It might disturb the service of individual service providers and it's expected to leave it to the markets.

And the target of this policy would be limited to those who are willing to request and are eligible to subsequent v4 allocations from the remaining pool. Whereas we have to get the whole Internet moving to the v6 to make it more effective. So the target of how it should work and the target of this policy applied is quite limited. So maybe it's not as effective as it intends it to be. And we might create barriers for those people who are just starting to implement v4 but they still don't have enough capitals or readiness in service to deploy v6. So is it wise to create barriers to those people, by asking them to deploy v6 and v4 at the same time? So that was the possible down side of this option. The second I mentioned to try to tackle the issue of those without a single global v4, do not have needs to connect to the v4 Internet. And this one - so let's give priorities to those who are without v4 at the moment which is finish your allocations. And the idea behind it is of course, yes, those people currently with v4 also have needs to request for subsequent v4 allocations. But since they already have a certain amount of v4 address space available, maybe they can use technology such as translators and then assign private v4 address space or v6 address space for the expanding needs and then NAT it to connect to the global v4. Those without one single v4 address space, they cannot do that. So they have no ways whatsoever to connect to the v4 Internet. And another option to help those, the expanding needs of the existing users could be that they can be on the waiting list to receive the recovered address space to meet the needs of the expansion, while for initial allocations it can be allocated from fresh RIR pool. And the advantage of this is at least we believe that it enables effective use of the same size of our address space. For example, a single /8 will enable initial allocations for definitely more than 10 years. It will last for more than 10 years if we distribute it to only single allocations. However, the same size, a single /8 will only last for three or four months if we distribute it based on the current criteria to any LIRs that come up with the requests. So wouldn't it be better if we make efficient use of the same space?

Of course, even this space will run out at some point or the other. But 10 years will probably give us enough time for the technology to develop, to ensure smooth communication between this. It would give us enough time to get prepared and make sure that those without the single v4 at the moment can also connect to the Internet while the technology gets ready until the point we no longer have to worry about IP versions. Those will be the advantages of implementing this option. However, the downside of this is those who will benefit from this option would be those who are only newcomers. Those receiving the initial allocations. So maybe some people will feel this isn't fair and this was, in fact, some of the comments made in our open policy meeting. Our existing LIRs felt their expansion, the needs for expansion is as important as the needs for the start-up business. So it could be quite controversial to weigh the importance of priority on who we should distribute the remaining pool too. So that's a possible downside.

The last option would be to distribute to public translators that do a translation between v4 and v6 networks. And the advantage of this is that it would probably benefit the Internet as a whole, not to just certain groups of people. And if enough numbers of translators are set up, then we don't really have to worry about the difference in IP versions. So they can receive v6 address space and can connect to v6 Internet and vice versa. But the downside of it is in reality would people really set up those translator networks just sort public? It might be likely that each ISPs will set up those translators for their own customers and infrastructure and if we give priorities to those, it will simply benefit that particular ISP and their customers and not the people in the Internet as a whole. So the downside remains the same as the initial allocation that the benefit will be limited to certain groups of people. And it night be quite difficult to, you know, speculate how much do we need to reserve for this purpose? It's quite difficult to define what the public translator network would be. So those could be the possible issues regarding this option. So I've given a summary on each measures and pros and cons. So maybe we can come back to this, like, later, in the discussion times so you will have a better summary picture of what the options are and what are the benefits and disadvantages.

And one additional issue that has recently become clear to us is that it seems that those ISPs providing IDC and hosting services would face quite serious problems. Because the exhaustion of v4 will directly affect their business model and let me explain how this is different from ISPs providing connection services to its customers. In case of ISPs can simply providing connection services on customer clients, they can assign private address space or v6 address space and then do the conversion. So it doesn't - the ratio of global v4 and the number of customers, it doesn't necessarily have to be one to one. Whereas the customers of IDCs and hosting service providers are all set-up servers. So the ratio of global address space, one global address space is definitely needed for a single server. So the number of customers expands, they definitely still need to assign a global v4 address space. So for them, not being able to receive subsequent allocations, expansions, that would mean that they can no longer expand the existing business and so we have found out that, well, we feel that they would face more serious problems compared to ISPs that simply provide connection services. Well, you might have different ideas, so please share with us later if you can think of different issues also for ISPs that only provides connection services. And JPNIC's position regarding these different options - as the very basic underlying principle, we think these issues that each region will be facing at the time of exhaustion widely vary. So for a region where the development of v4 is pretty much done and most of the ISPs in the region have already received v4 address space, maybe the most important issue for them is to encourage v6 deployment. So whereas in a region where v4 Internet still has rooms for development and not many ISPs have v4 address space available, maybe the priority would be to make sure that there will be v4 address space available to as many members of the community as much as possible. So maybe it's not so wise to deploy exactly the same policy throughout the region. So we think tolerating regional diversity will be quite important.

So based on this basic idea, we think that for the use of the remaining last piece of APNIC pool, we think it would be better to reserve it for a special purpose, rather than just distributing everything out under the current criteria. For example, maybe give priority to initial allocations. Give priority to public translators to help address the issue of those without v4 don't have means to connect to the v4 Internet. And this is why we came up with a global policy proposal and it links to this idea. And I will explain in more details how this will be connected in my next presentation. So, a question I have to the community is, "What do you think was the most important issue which should be addressed in the APNIC region, looking at the situation of the Internet development in that region, as well as the v4 address distribution that has been made so far?"

And I've listed the issues here - so should we help those without any v4, should this be the most important priority or should we ensure smooth communication between native v4 and v6 or encourage v6 deployment, etc. And we'd like to consider if there is any policy that would help to encourage the most important issue you consider in our region.

And of course these are just the options we came up with so far, so if there are any ideas or possible issues that we haven't really come up with, other suggestions, are certainly very welcome.

So while facing the exhaustion instead of waiting, let's be proactive. Thank you. APPLAUSE

PAUL WILSON:

Izumi, Paul Wilson from APNIC. That was excellent. Thank you. Can you hear me? A very interesting way to present that set of problems and to look at a methodic breakdown of options. Thank you very much.

I may be overlooking something but I think there's a class of issues that isn't addressed here because I think what you're talking about is the management of IPv4 address space up until the exhaustion of the IANA and RIR free pools. Then of course after that we're still managing IPv4 address space and there's a whole class of policy issues which come after that date. How exactly do RIRs behave, what are the policies and all of the different options to how IPv4 can be managed? I think we should be talking about those now because it would be nice to think we can put that date back by 10 years but if it is as close as it might be, it might be very good to be preparing ourselves for the post-exhaustion.

IZUMI OKUTANI:

Thank you very much. We expressly did not include that but I agree that it's an important issue to be discussed and, yes, it's really welcome to hear comments about what we should do after the exhaustion as well.

ED LEWIS:

Up to now, I haven't thought much about... maybe I'll use two microphones at once. Can you hear, Randy? Up until now, I haven't given much thought about who gets the last address and I thought it was to don't give it to someone who wants to horde it. After hearing some talks earlier today and this week was what's the transition from IPv4 to IPv6, dual stack looks like the only workable solution. That's what I got from the other presentations. We need to look at where we put the last v4s so we can't just give it to the people who will hold it. To get on to v6, you have to have v4. It's an issue for v4. It's an issue to start thinking about.

GEOFF HUSTON:

Geoff Huston, APNIC. Let's put a bit of this transition in and try and understand the leverage you've got.

What we understand from the numbers so far is that 46 /8s on current rates of growth in v4 will fuel the Internet for about the next 30 months. Now, the dual stack model is quite a protracted process and if you actually understand the way industry works, and the scant margins of revenue from existing deployments, then no-one is going to rush out and do dual stack because no-one has infinite money. So the one thing people can expect is the demand for v4 in dual stack because every single new deployment needs to be dual stacked, is going to last for not two years, 30 /8s, but over 10. How many /8s is that? Well, we're going to burn through 46 /8s in the next 30 months. So in the next 300 months, that's 460 /8s. Yep? I'm doing very simple maths here to illustrate the fact that if you're trying to fuel the next 10 years of Internet growth in dual stack, the amount of v4 address that is required in the current environment of the way we deploy v4 is purely massive.

And the leverage you've got with the last few /8s sitting in these pools is close to nothing. It's the transition issue which is dominating most of this and trying to understand how you can play with the very, very small amount of resource you've got left when faced with the very large demand that you haven't. We haven't got any capability to supply on, is really the issue here. We don't understand what translators are. At the moment, they're just mythware. They are Mythware. We don't understand this. There's always going to be initial allocations all the way around. The network will continue to grow and the real issue is at some point the line stops in v4 and you can't service it and the network will continue to grow and trying to service that is the big issue. You can't make 46 /8s do a decade work of magic in dual stack, that math just doesn't work. And somehow it's not really these residual /8s that really count.

IZUMI OKUTANI:

I agree with the point Geoff just mentioned. We don't really have a clear picture of what translators will be but the basic underlying idea behind the translators, not to allocate to each individual networks to allow dual stack, but have like, I don't know, a translator network that allows connections of each individual v4 networks and transform to v4 and vice versa. That's the idea we had in mind. And that way I think the requirements will be much more restricted than allocating v4 to all the eight individual expanding networks, at least dual stack.

JORDI PALET:

I think one of the issues that we are missing in this picture is that people understand dual stack as public v4 address and that's wrong. It's not the best choice. It will be much better that everyone at home has hundreds of public IPv4 addresses but that's not the reality. And the same way we have situations where there are different levels of NAT today, we will be in the same situation for the next few years until everyone or almost everyone has done the transition to IPv6. Again, it's not an optimal situation, but we can live with that. So trying to somehow extend the life of IPv4 or trying to provide IPv4 addresses to the registries as this policy proposal is trying, I think it's not really providing anything good. I think in the other way, trying to buy time for something we don't really have. I think it's much better to keep in mind we can do the transition for addresses, let's go for it.

GEOFF HUSTON:

Geoff Huston, APNIC. Let me respond to this one question about can you use private IPv4 addresses in the context of moving packets around in a public IPv4 network in a dual stack. We can't do it today or tomorrow. By the time the packet hits the routing centre, it has public addresses in the destination field. And for the packet to be returned from the other side, the return packet needs a public address in its destination field. The inference is no matter how you do it or how many NATs you have at either end, in the middle you need public addresses. The mythology that somehow NATs can do something they never did today but can do something magically tomorrow is another case of mythology.

IZUMI OKUTANI:

Let me clarify this is just discussions and we're not intending to prolong the IPv4 address space in any way. It's just as a clarification to your comment, Jordi.

JORDI PALET:

Geoff, I agree with what you said. I'm not saying using private addresses on the public Internet. I'm saying we may increase the levels of NAT. We might need to cap private addresses at some point but of course every ISP need to have at least one public IPv4 address but that will work even if just one public IPv4 address is available for every ISP, that will work.

BILL MANNING:

Is that better. So I'm looking at those possible ideas up there and I don't think we need to encourage IPv6 deployment. I think things will happen so that IPv6 will be seen as perhaps the only alternative on the way forward. Efficient use will happen even if we do nothing. So I suspect that the biggest issue would be priority to translators, it's not NATs. IPv6 is a different address family. This is like IPv4. They do have to be translators, so engineering efforts to link those two together probably should be a high priority and whether or not there needs to be policy for that, I don't think so.

IZUMI OKUTANI:

Thank you.

RANDY BUSH:

I would like to just stress one thing. My opinion is that our job in this community is to make the Internet work for our members and for everybody in the world as best we can. And these decisions are micro-decisions about some technology choices about how. But our job is not to make NATs wonderful, not to make v6 wonderful, not to make v4 wonderful, it's to make Paul's grandmother able to talk to her grand-daughter.

YAN MA:

My comment is that we should expect more discussion on the migration issue. One comment is that I agree to support your suggestion and I wanted to see more and more of those kind of in-depth discussions and proposals to be held at the next APNIC meeting. And I'd like to know is there any collaboration with any other RIRs related to this issue or not?

RANDY BUSH:

I warned you you'd already heard too much from me. This proposal in its various forms is being presented in all the RIRs. And there's much discussion and the people making the proposal have been very responsive to that discussion and indeed the form of it, when you see the actual proposal, is kind of requires it being global, which is a very difficult thing in the five RIR communities. Does that answer your question?

IZUMI OKUTANI:

This part is a discussion paper. Not policy proposal. What we're trying to do is clearly from the floor here is to try and come up with a policy proposal at the next meeting if people think we should have some kind of policy proposal to help some of the issues.

PAUL WILSON:

Paul Wilson again from APNIC. Can you hear me with this? Can you hear me with this! OK, thank you. There is a lot of interchange amongst the RIRs at both the level of attendees to the meetings, the address council and to some extent RIR staff members who visit other RIR meetings and help to improve information exchange in mutual understanding amongst the RIRs. That's something which has been going on for quite some time. I think it's accelerating this. Each of the RIRs is an independent community. And the ability and the motivation amongst RIRs to coordinate and to have consistent policies across the board is the motivation is there but I think the autonomy of each RIR is a fairly sacred issue. The need for a global policy in this area, I think, is that in order to provide an RIR with the freedom to implement the types of measures that we're talking about here, there actually needs to be an environment in which they have some ability to make those decisions and to change the way addresses are being managed within their region without adding, for instance, the rest of the addresses consumed by the other RIRs.

For instance - I'm sorry if it's not entirely clear - but if the APNIC community wished to implement a policy here that would throttle down the rate of consumption of addresses in this community, then unless that is reciprocated elsewhere, then the addresses can keep being consumed at a global level and at the end of the process, APNIC, the APNIC region is left with a smaller amount of addresses and IPv4 addresses are still entirely consumed. It would be quite a self-defeating thing in that environment for us to decide to throttle down the address space. That is something to face. We do have freedom and we insist on our freedoms but there is a necessary inter-dependency. This is where it's very important and we have to effectively exercise the autonomy and ability to determine our own future. We want to do that in an environment where we're not cutting off our noses despite our face, which is a phrase Randy used yesterday in the fee discussion. A proposal a region has brought forward in allowing RIRs to meaningfully determine their own fate. I think we need to recognise the interplay of what could be quite simple and straight forward, global policy proposals and our ability to actually make decisions that are at a sophisticated level on a regional basis. And such decisions I think would be impossible on a global level.

IZUMI OKUTANI:

In case it wasn't clear, this part is intended to be a regional proposal because it involves the distribution of APNIC pool. And in my next presentation, I'll be discussing a global policy which discusses about the distribution of IANA RIRs. So to an certain extent we need to be aware of the situation of other RIRs but we can probably decide what's the best within our region, we don't really have to worry so much about making sure other RIRs implement the same policy in this area. Would it be possible to get a feeling from the floor, which options, which of the five options that people support from this? It's not proposal but in order for us to make a proposal next time, I'd like to have a feeling from the participants here if there is a particular option out of these five that people support in general. Would that be possible?

SAMREEN FATIMA:

I have a few words - I am from Pakistan. This is the first time for me at APNIC. This is the first time I'm viewing the procedure how the policies are formed at APNIC. The feeling that I've got is that we're focusing on the ways, how to transition from IPv4 to IPv6. And the technicalities of the transition and the technicalities that the IPv4 pool is depleting and we have to in one way or other move to IPv6. But we all are, I think not focusing on one small issue that this transition or completely shifting on IPv6 would involve the management part of the ISPs who are not so technically knowing about this situation is depleting. I would very like this idea of yours to lay down some sort of policy of APNIC how to distribute their remaining IPv4. I would also like to add on a little bit.

For every new IPv4 allocation, whether it be an initial allocation or a translator or the other ISPs who already are a member of APNIC and have IPv4, to make it mandatory for them to have some part of their traffic over IPv6, so that they are sort of forced to start thinking about IPv6 and start enlisting on the transition process of moving from IPv4 to IPv6. So I would like to add that we can choose any combination of the top four options but what would motivate the management to choose any of the four?

IZUMI OKUTANI:

Thank you. You're suggesting that it's not that we can only choose one but the combinations of these options are possible. Is that your point?

SAMREEN FATIMA:

And what I'm saying is that if we are trying to make it a policy of APNIC, we should add on that for every new IPv4 allocation, part of the traffic over Internet should be on IPv6. So that they have to adopt to IPv6, but whether the new allocations are whether they already have their own IPv4 allocations.

RANDY BUSH:

Excuse me, I need to interrupt. We're now five minutes in to lunch. The chairs have talked and we kind of wanted to do something differently this time. These are critical issues and we wanted to let the proposals be presented and then because of people's language difficulty, to allow lunch breaks and coffee or tea breaks, pardon me, for people with common language to talk together about the issues. So may I suggest that lunch was supposed to be from 1 o'clock to 2:30 but instead we have it from 1:30 to 2:30. So we can finish this and have the proposal so you have the opportunity to discuss it at lunch. Does anybody object to that? Thank you.

IZUMI OKUTANI:

I'm fine with that. OK. I'm just trying to change the slides. Sorry. It's taking a bit of time. Sorry about this. I'm going to try again. I'll give it one more try and if it doesn't work... no, it's not working. Oh! Very sorry to have kept you all waiting.

So this is the second part of my presentation. And this is the global policy proposal. And in my earlier presentation, which was informational, we discussed about the policy distribution of the remaining APNIC pool at the time of v4 exhaustion and this is trying to discuss how we should distribute the remaining IANAs to RIRs. It's a revised version of a proposal we made in APNIC 23 and it's intended to reduce, avoid confusion at the time of exhaustion by taking two measures. Firstly, clearly defining how we should distribute the last pieces of v4 address block and the second measure would be to keep the community well-informed of the situation. And our possible issues in terms of address management perspective - at the time of exhaustion is there's no clear agreement on how we should handle the last piece of IPv4.

For neither for IANA to RIR allocations or RIR to LIR allocations. It could be possible we should carry out to the end, by the current policy. But we should explicitly make the decision we should carry it out until the end under the current criteria. And we haven't really made agreement about this. And the second point is that there has been projections made by personal efforts and which are very useful but no official projections or information provided by RIRs on when would be the last timing that LIRs can request for IPv4 allocations from their respective RIRs. So it would make it a little bit difficult for LIRs to plan ahead.

And the third point is what I mentioned in my first presentation, the issues facing at the time of exhaustion vary depending on each LIR region. Maybe a certain RIR region feels it would be more important than the others. While the others are ensuring that v4 will be equally distributed. It might vary. Those are the possible issues we came up with at the time of exhaustion.

This is a brief introduction - next. This is what we proposed in our last meeting and I don't really want to go in to much detail about what it was because it was already in the past. But the basic idea we proposed was let's artificially stop allocations of IANA to RIRs after a certain date. For example, when there is only 30 /8s available at IANA, let's stop allocations and then this will be the date that each LIRs can request for additional v4 space to RIRs. Let's announce this. This was the idea we suggested in our first version. And we must admit that this wasn't exactly very popular among other RIRs and strong concern was expressed over legal implications of this policy, that even though there are three remaining pools available within RIRs, and by stopping the allocations and making them not available to the others, it would have legal problems because it would seem that RIRs are taking advantage of its monopoly and not giving the necessary space to those who need it. So no consensus was reached for this proposal outside APNIC region. So we made a revision and this time we decided that we should ensure that all blocks will be distributed, carried out, until the very end to address the concern - we shouldn't artificially reserve any v4 blocks. We also realised the distribution policy for IANA to RIRs and RIRs to LIRs, they are different. So they should be defined separately. So this proposal describes the distribution policy of IANA to RIRs. My earlier information presentation was to discuss the distribution policy from APNIC to LIRs. And we also feel that in policy, maybe this is a - it should be explicitly clarified in sharing with everyone that tolerance of regional diversity should be allowed in the distribution of each RIR's pool to LIR's. We thought it was important to continue in keeping community informed of the situation.

This is still an important issue. So we try to include this aspect in to our proposal this time. So the details of the proposal is that firstly, we think we should distribute the remaining IANA pool equally to each RIR. And our idea is that when three IANA pool available hits - five times /8 - then we should distribute a single /8 each to RIRs and we define this as IANA exhaustion date. The second element of our proposal is RIRs should maintain the current distribution criteria, do not change the policy until this IED. And after IED, we feel that it should be made clear that each RIRs can define their own policy which best fits the situation of their region. And the possible options are the ones I mentioned in my earlier information presentation. So those are the possible options we explored for our region. And if some RIRs feel there's no need to change the policy, let's stick with the current distribution, that's totally fair enough.

But those decisions should be allowed to make for each individual RIRs. And the fourth point is that RIRs should provide official projection of IED. Of course, there are projections are readily available by personal efforts and they're very useful, but I think it makes a huge difference whether RIRs officially makes and provides this information are not to make the community aware. So they should take opportunity to public meetings or any other effective means to keep the community as informed as much as possible of the situation.

So how would this help? Firstly, it would help define distribution of the last pieces of IANA blocks to RIRs in advance. And it would also help RIRs to project to expect what would be, when would be the last time that each RIR will be exhausted. By equally distributing the same size in advance to each RIRs. This would allow each RIR community to define their own community to match the situation of their own region. And LIRs will be better informed of the situation of the exhaustion. So I've summarised. Which of the proposal elements would help the issues I've listed? And you can probably see it for yourself, so I wouldn't bother reading them out. And that's basically it.

RANDY BUSH:

Thank you. Can I try to clarify - there are three essential changes your proposing. The first change is when there are only five /8s left, one goes to each RIR. Secondly, and I don't think you intend this, that you're telling the RIRs they must maintain their current criteria until the IED. When, in fact, probably some flexibility there is desirable. And third is that you kind of expect them to make serious projections.

IZUMI OKUTANI:

There can be projections. So, yes.

RANDY BUSH:

So those are the three, you know changes.

IZUMI OKUTANI:

Yes.

RANDY BUSH:

Because doing your own policy after IED will change because people do their own policies.

IZUMI OKUTANI:

Thank you for the clarification. It's right.

RANDY BUSH:

So we have 12 minutes left. How should we spend them? Should we spend them hearing from Haitham?

Prop-051 Global Policy for the Allocation of the Remaining IPv4 Address Space

HAITHAM EL NAKHAL:

Good afternoon, ladies and gentlemen. So now we come to the most interesting policy. I know you have been waiting for it. It's the global policy for the allocation of the remaining IPv4 address space. I would like to present the policy proposal 51. IPv4 allocation policy to RIR. Now, I am using the demand policy to allocation IPv4, which states that RIR is allowed to apply for another allocation if it has less than 50% in its pool and there is not enough to cover it in the next six months. Nowadays, they run a gentleman's agreement.

So the proposed policy incentives: IANA free pool for allocation of IPv4 addresses, /8s, is decreasing rapidly. It brings certainty to each RIR that they will receive a last IPv4 allocation from IANA of equal size.

Policy statement. Policy is divided into two phases. Phase one - IANA reference N /8 units to each RIR. IANA keeps applying the current allocation policy until the request for IPv4 from any RIR to IANA will compromise the remaining free pool of IANA according to the following formula - this formula is not existing in the proposal text, but I put it here just for explanation.

If X = IPv4 /8 units available before the last request, and A = /8 units needed to fulfil the last request from an RIR, if R = the number of RIRs recognised by ICANN, then, if R x N < = (X - A) that is the threshold value.

Then, at this time, phase two of this policy will be initiated.

So what's phase two? IANA automatically allocates the reserved IPv4 allocation, N units, to each RIR. And responds to the last request with the remaining available allocation units in IANA pool, which is M units.

So what's M units? Calculation of the remaining M units - assignment for each RIR equals reserved N /8 units, so the remaining M units will be the available /8 IPv4 units before the last request, minus N multiplied by R.

So the total assigned /8 units for the last requesting RIR equals N plus M.

Again, this equation, this formula, is not in the proposal text, OK.

OK, so now, what's the N value? After the discussions that took place on the mailing list, we are suggesting that M = 2. Why M = 2? Today, IANA allocates two /8s according to the gentlemen agreement to any requesting RIR, so the proposed allocation will have the same size as today's allocation. With two /8s, each RIR will have an allocation big enough to enable developing of more Conservative LIR allocation policies. With N equal two we can say we are not boosting RIR shopping. It's not a big enough pool to do that.

As a quick example for what we talk about, assume the remaining free pool for IANA equals 11 and an RIR requests for two /8 IPv4. Then IANA will allocate N for each RIR and in addition allocate M equals one to the last requesting RIR. So the total /8s allocated to that last RIR is two plus one.

So the proposal advantages - it allows each RIR to guarantee its last allocation units so that each RIR community can develop its own mechanism and policy for making use of that last IPv4 allocation from now.

If the equal allocation of the final /8 blocks across RIRs brings certainty that all RIRs will have a final allocation from IANA. And it limits RIR shopping.

Are reduces pressure on IANA central pool. It allows for suitable time for LIRs to begin their transition phase to the next IP generation, which is IP version 6. Provide real IP version 4 for newcomers, new projects, to avoid using NATs at the beginning, as many applications encountered problems while using NAT.

I have put in a link for newcomers and new projects.

The proposal status. It's submitted for AfriNIC in July 2007 and for APNIC in July and it was discussed on the mailing list and is now in the face-to-face meeting. For ARIN, submitted in July 2007 and for LACNIC, somehow, the proposal was not known of the existence of the proposal by Izumi, and it has consensus in the LACNIC meeting and for ARIN it's submitted in July 2007. So thank you.

As an example for newcomers, we couldn't say for newcomers for three years or so, we don't have I'm version four for you, go and use IPv6 or you will do. They could live in isolated island and so on, and so, as no-one can predict exactly when IP version six - oops - we couldn't say for newcomers after the following three years or so, "We don't have IPv4 for you. Go and us IP v6 or you will die." They could live in isolated island as no-one can predict exactly when IPv6 will be dominated. And for existing LIRs, each RIR community can develop policies for them for encouraging them to deploy IPv6 for requesting the new allocation of IPv4, as Izumi said before. It's the role of RIR community to plan how to afford IPv4 in the next years and this proposal is aiming to help in that.

So for an example for this, about newcomers, so for an example of projects, as you will see, Africa is divided into five regions - north, east, west, south, and central. Green cables are planned and funded. And dark green cables, here in the south, are needed. Blue cables are required. Red and orange are the proposed submarine cable for a new project in Africa.

So - I am sorry. I make a hyperlink but it doesn't work, OK - for the newcomers and for new projects. OK.

So this cable is - this cabling project called NEPAD - the New Partnership for Africa's Development, has identified that ICT infrastructure is one of the major parameters for economic growth and poverty reduction in Africa. Africa needs a broadband infrastructure linking all its nations. Africa needs to extend this infrastructure within the individual nations in a manner that permits modern computer-controlled networks to function in support of all elements of modern economy.

So, as a plan for using this cable, for these cables, two of the four IGF streams are access and diversity. - themes are access and diversity. So on the access theme, as a plan for using these continental fibre cables in developing ICT in Africa, there is a plan for establishing, five regional Internet hub, four regional Internet carriers, three continental Internet carriers, transit IP traffic.

I mentioned these two projects - newcomers and a definition for why we have - we need for - to reserve two /8s for future planning. You know, we are planning for the coming five years, so we would like to know from now, how much we have in the next five years.

So this is the reason why we're asking for the reserve of two /8s for each RIR.

That's all.

Discussion about 'IPv4 countdown policy' and 'Global policy for the allocation of the remaining IPv4 address space'

RANDY BUSH:

Procedurally, we have two proposals here - the one Izumi-san presented and the one Haitham presented.

Am I correct that essentially the difference is really that you want to reserve two /8s and the proposal from JPNIC and others, which Izumi-san said, was one /8 per RIR?

Also more detail is Izumi-san presented some discussion of suggested allocation strategy before - allocation policy before and after, not radical changes, but saying we keep going up until then, but after, that each region actually will set its own policy, etc. Haitham is not saying how policies would be set before or after.

HAITHAM EL NAKHAL:

After - it's in the FAQ.

RANDY BUSH:

Pardon?

HAITHAM EL NAKHAL:

I prepared for a question so I gathered some questions and I am fixing on the mailing list, so I gather it and put it in a FAQ, so questions and answers for debate on how to make...

RANDY BUSH:

OK.

Um, is there any critical question for these two proposals before we break for lunch and are able to discuss it? Thank you, Mr Plzak.

RAY PLZAK:

It's not a question. Ray Plzak, ARIN. It's not a question. It's an observation. The statement was made that the current IANA policy is to allocate two /8s to each RIR. That is not the policy. The policy is the RIRs will request what they need and the RIRs have agreed that, despite what they need, if it's more than two, they will only accept two, which means that, if an RIR needs less than two, they will not get two, OK? And that's not what you said. You said the policy was changed and that the IANA was allocating to the RIRs two /8s. That's not true.

HAITHAM EL NAKHAL:

OK.

RANDY BUSH:

So, Ray, hidden under your point is that this - that Izumi-san's proposal, which allocates the last five as one each, would actually require change of policy in the relationship to the IANA, because they'd be requesting less than two.

RAY PLZAK:

The current criteria, global policy criteria, is established such that the RIRs will request and will be given what they need, OK? Now, the question is, is that what happens when the pool decreases below that what is needed by an RIR? What do you do? And so there are several options there, OK? The point is that every RIR knows that at some point in time they're not going to get any more address space from the IANA, so none of these proposals are adding anything to that. We all know that.

So the point is what do you do when the IANA can no longer fulfil the need of an RIR? The same problem the RIR is going to have with an ISP or LIR, is that at some point they're not going to be able to fulfil. They're going to have to say, "No, I don't have it." And so the question becomes, do you hit a time when you artificially say, "No, I don't have it," or do you just let it happen. That's not the issue. I did not come here to discuss the merits of the policy. I was trying to correct what I heard as being an incorrect statement of the current policy.

RANDY BUSH:

Thank you, Ray.

Ah, I thought you were walking to the mic, Geoff.

GEOFF HUSTON:

No.

RANDY BUSH:

Eat food now.

(End of session)

Policy SIG - Session Two

Thursday 6 September 2007

1430-1600

RANDY BUSH:

You may have been wondering all this time why Anne and Sanjaya are up here. We have been thinking of trying an experiment that, instead of having the transcription up there, that they have volunteered to, very simply as possible, phrase the discussion, instead of a transcription of the discussion, to have the idea that is being discussed visible. We're trying to understand if we have ways to make it easier for people whose native language is not of the speaker.

Does anybody wish to express a preference for either seeing the transcription, like you are seeing now, or seeing a summary?

The transcription will still be made and will be available later and so will the movies. And it's online right now. You can watch the transcription online right now. Versus having succinct summary of what the discussion is on that screen. Would anybody like to make a preference?

That's exciting.

GEOFF HUSTON:

The summary.

RANDY BUSH:

Does anybody want to go back to lunch instead of this horrible policy stuff? OK. We'll try the summary.

So does somebody understand the technology?

Two things - number one is does somebody understand the technology to make your summary visible?

SANJAYA:

Do you want it shown there or shown there? This one is connected to that one.

RANDY BUSH:

Well, what's the presenter connected to?

SANJAYA:

You want it shown here?

RANDY BUSH:

I don't care. I just care that the presenter's on one and you are on the other.

And secondly is the URL for the transcript well known?

And maybe that is the first bit of summary you should put up.

It's right at the top of the webpage for the agenda for this meeting. It says 'read transcript'.

OK, Toshi has just given me the privilege of some housekeeping announcements.

Again, if you have questions or answers, please use the microphone. Please state your name, because of the transcription and people online who may be only here. And this session is broadcast and recorded in video and all those things.

From 1800 to 1900 this evening is the MyAPNIC BoF in the Crystal Ballroom for future and current members of APNIC. APNIC would like to hear about your experiences with the MyAPNIC portal and discuss how they could make the portal more beneficial to you.

There will be Lightning Talks from 1800 to 1900 in Regal 1. That's downstairs. Lightning Talks are informal, very short presentations on very current things.

The APNIC Policy Flash demo are all day at the services lounge and Helpdesk, which is hidden where? Downstairs again?

And there's a feedback form online on the website under the 'local information' category. To encourage you to fill out the feedback form, there is a prize, which is an MP3 player. So submit your feedback form before 1600 at the end of tomorrow, and you will be eligible for the prize, which will be drawn at 1600 tomorrow. And the prize-winner will need to be in the room. So no submitting your feedback form and running away. You actually have to be there to get the prize.

The Closing Reception is tonight in the Terrace Plaza next to the Crystal Ballroom. The reception starts at 2000 this evening.

Now, I think Toshi-san has the agenda discussion.

TOSHIYUKI HOSAKA:

Thank you, Randy. Since we have not much time for the discussion of the policy proposals, so I'd like to have some agenda-bashing for the rest of the day, which I have, I should have done at the beginning of the Policy SIG.

But, anyway, after this agenda bashing, we will resume the discussion on the proposal from Izumi and Haitham. And after that, we will discuss APNIC Transfer policy from Geoff Huston. And after that, we are discussing, we are going to discuss the global policy for the allocation of AS numbers from IANA to RIR. And next we have the agenda from Jordi - IPv6-ULA-central. But I heard that Jordi is not going to make this presentation. Is that correct, Jordi? Are you here? OK.

Then we can drop this item from the agenda.

And next one is proposal 043 - proposal to remove reference to eye six policy document as an 'interim' policy document. This is the proposal which was presented in last APNIC meeting and we sent back - we have sent back this proposal to the mailing list for further discussion, but no discussion has been occurring in the mailing list, so I would like to - I also would like to drop this proposal from the agenda. Jordi, do you agree? OK. Thank you. So, again, from Jordi, we have a proposal to change IPv6 initial allocation criteria, proposal number 042. This proposal also has been discussed in last APNIC meeting and sent back to the mailing list for further discussion, but no discussion has been occurred in the mailing list, so I would like to drop this one, if Jordi agrees. Do you agree? OK. So we will drop this one from today's agenda too.

So the remaining presentation is all informational ones, so if we have enough time to present these proposals, we will do that, but rather than keep time for these informational discussions, we would like to have much more time for discussion, so if we have time, we will go through these informational presentations.

Thank you very much for your understanding.

RANDY BUSH:

So now we'll go back to the proposals for the IPv4 run-out and that's proposal 46, which Izumi-san has made and proposal 51, which Nakhal has made. And, first, if you don't mind a slight review is the core of the both proposals is the last of the IPv4 space to be given evenly to the five registries. Nakhal says two each. Okutani-san says one each. Okutani-san's proposal from JPNIC has a little more data about what the policies would be before and after.

Are there questions? Or discussion? On those two run-out proposals?

LEO VEGODA:

Hello. My name is Leo Vegoda from IANA.

I've got a question for, perhaps both proposers. Both proposals call for a numerical output, so each RIR would receive the same amount of address space. Other than looking at the allocation history, what were the numerical inputs that went into the calculations that came out with identical outcomes for each RIR?

RANDY BUSH:

Let me be tactless, Leo. What you are saying is, since the consumption rates of the RIRs differ so widely, how come both proposals allocate evenly?

LEO VEGODA:

I was hoping to -

RANDY BUSH:

Be more subtle?

LEO VEGODA:

Well, I wanted to take more into consideration than just consumption rates. Because it's something that there's lots of different factors, not just the consumption rates, but the number of people, the rate of development in different countries. And I was just wondering which of these factors have been considered when coming up with those numbers.

HAITHAM EL NAKHAL:

Hello. So you would like to know what's impact of this proposal at APNIC or AfriNIC or LACNIC or all RIRs? Am I right?

RANDY BUSH:

No. There are different factors - one is the population, one is consumption rate, another is latitude. What are the factors that were considered that caused you to come to the conclusion of a flat distribution?

HAITHAM EL NAKHAL:

What are you asking about?

This policy has no negative impact on the APNIC region and on the other hand it gives more certainty to the regional community. The percentage of the number of incompetent addresses allocated to APNIC in relation to the number of Internet users is much smaller than in other regions. In fact, the percentage are closer to LACNIC and AfriNIC than in North America. It means that the challenge that APNIC region faces is not very different from other regions.

If this policy is not approved, APNIC can be the second in the queue when the last two /8s are ready to be allocated, which means that you have addressed for satisfying the needs of your customers for less than six months. APNIC will be, so, the first running out and the situation in APNIC region may become more worse than in other regions. As the wasting of addresses in the APNIC region is not very big. So APNIC may have problems for optimising the use of IPv4 addresses.

How much addresses are not being used in APNIC region, I don't know. And how much addresses could potentially recover. Of course, you know.

On the other hand, if we apply this proposal, it will permit implementation of soft landing policies in advance, because you will not have to compete for the last allocation and at the end, you will receive two /8s which will permit APNIC to develop suitable policies for living with the IPv4 decaying in a known base. So there is no negative impact but, in contrast, there is a positive impact.

RANDY BUSH:

Izumi-san, perchance could you answer Leo's question?

IZUMI OKUTANI:

I think you want - we did consider an option of just carrying it out without distributing equally to each RIRs and, you know, just to distribute on actual consumption basis versus distributing a single /8 to each RIR. And what we want to achieve through this equal distribution is that, if we don't do that for a single /8, RIRs wouldn't know until the very last minute if this remaining /8 will come to their pool or other RIRs until, you know, the very last minute. So once it's - it would make it more easier to speculate the remaining pool within their region and for each RIRs to, you know, notify the LIRs. OK, this is the remaining pool that we expect to have. So, until this date, you will be able to request more v4 requests.

And then the second point is related to my informational presentation. We think that it would be useful to reserve, for example, a single /8 to a special purpose rather than just distributing out to everybody. And, again, if we don't know until the last minute whether we're going to have an additional /2 or /1 or, you know, you might not get any, you know, depending on the timing, it makes it really difficult to, you know, keep this - how much block you should keep for this purpose. So we think that it would make things more - I don't know, easier to expecting - know what to expect.

And then we also thought about its impact that, by not basing if on the actual consumption, would it fasten the exhaustion rate of a certain region, and then, according to our calculation, I think RIPE, ARIN and APNIC, they consume maybe three to four /8s per year and AfriNIC - correct me if I'm wrong - but maybe it takes two years to consume auto/8 and AfriNIC, one /8 for one year. Something like that. So distributing a single /8 to each region, it wouldn't really fasten the consumption for APNIC, RIPE and LACNIC that much, because it would be maybe two, three months. It might. But then depending on the timing, it's also possible that they would have gotten zero allocations, so, in that case, I think the actual impact is pretty much small and negligible and we thought that the benefit outweighs the impact that it has and that's why we came up with this idea. I hope it answers your question.

LEO VEGODA:

Yeah. Fair point.

RANDY BUSH:

Let me see if I can just take the critical point, which is remove the surprise so that we have not got five RIRs coming to the last minute and only three of them getting. Everybody knows that, as we get to the end, you will get a small lunch, so that you will get a small lunch, OK.

Haitham, I did not hear specifically what your criteria were for choosing flat and two, versus - I've seen the presentation and the proposals to death. Could you answer in English - two questions.

Why flat? Same as Izumi-san? Same reason? OK. No other magic physics. Fine.

Number two is why two instead of one?

HAITHAM EL NAKHAL:

OK, so we can split our two sections, one for newcomers and existing LIRs, so you can reserve a /8 for them, for the existing LIRs, you can put them to IPv6 before getting more IPv4. And we can't leave newcomers without IPv4. And a /8 can be reserved for continental projects.

RANDY BUSH:

You can do that with two /9s. My-san tried to express that when you're down at the last /8, you will have special policies, that is one you want. Other people in other regions will want others to be able to provide. So could you possibly compromise with Izumi-san and agree on one /8? We'll leave that. Next.

GERMAN VALDEZ:

I'm German Valdez from LACNIC. In the fact that we are discussing a global policy and I would like to share some information that is happening right now in Latin America, which is my region. And there is a lot of concern about the IPv4 deprecation and this was the trigger that we take the policy proposal that Haitham already showed in this session. And the fact is that the LACNIC board of directors received a lot of requests for guidelines regarding this policy, what is going to happen is going to happen once the IPv4 space is over. A lot of the discussion before this LACNIC meeting where this policy was approved and there was also discussion in other regions and the LACNIC board of directors feels and thinks that this is the correct approach, to get this problem.

I don't want to - as Paul Wilson said before, the autonomy of each RIR is sacred, so what we want to do is only share with you what is happening in LACNIC and in AfriNIC is the same thing, in this situation, and it's being developed in regions, the various concerns about is there going to be enough IPv4 for the last years of the proposal.

So that's it. Thanks.

RAY PLZAK:

I'm waiting for the chair to recognise me.

RANDY BUSH:

You're Ray Plzak. I recognise you.

RAY PLZAK:

You were watching your e-mail, Randy. I couldn't tell.

I want to comment on the native RIR practices, one of which is that the RIRs wake up in a brand new world where we're out of IPv4 space. Excuse me. The RIRs are continually forecasting the need for address space. They do it with their customers and they do it with the IANA.

So the RIRs always have generally a good idea of how much address space they have and how much they need and based upon conversations with our customers, how much they think they're going to need going forward in the future and they have to have this information in order to prepare in advance the allocation request to the IANA.

And so that is a fact. It wouldn't be all of a sudden, "We don't have any." Believe me, everybody watches what goes on. As far as figuring out when it's going to be when you won't get any, that goes back to what Geoff was talking about this morning, is that the closer you get to the end, the easier it is to predict when the end is going to occur. Right now, it's an iffy thing. Geoff can apply real fancy math to it and come one a precise date. Good, but it doesn't mean anything. He told you that as well, because it changes every day.

So I think that it would be useful if you don't couch this in the fact that you're doing the RIRs a favour because they are continually forecasting, as they do anyway. So that's one thing.

The second thing about the two /8s that are given to the RIRs. They're not given to the RIRs. The RIRs request address space. Now, what that means is, if I'm going to consume four of them in a year and I qualify for four when I come in the day and say, "I qualify for four," and IANA says, " You qualify for four." And I say, "No, just give me two," I guarantee I will be back in six months to ask for two more. I will continue to consume at the same rate, but I take it out in smaller chunks.

Oh, by the way, I'm not sure if we worked out a process on this but it seems to be that this is generating twice as much work for the IANA by the RIRs doing it because of the requests. So maybe we should go back to the gentleman's agreement, back to where we were and let it roll.

RANDY BUSH:

Ray, just to check, so are you suggesting that neither proposal is useful to you?

RAY PLZAK:

No, what I'm suggesting -

RANDY BUSH:

That was my question.

RAY PLZAK:

No, what I'm suggesting is that the basis for these proposals is not sound.

ROB SEASTROM:

Rob Seastrom, ARIN AC.

Can I get the slide that was previously up there before.

Apparently not.

The second point that was made - and I'll only speak to that second point - was that this policy of assigning two /8s would provide more time to come up with a good plan for conserving IP address space - IPv4 address space. And my question is what is keeping us from coming up with those policies now?

Why do we have to wait until we only have two /8s left? And what difference does it make whether it's two, one or whatever? At that point, the building is already on fire and we should be running for the door.

HAITHAM EL NAKHAL:

Can I answer now?

ROB SEASTROM:

Certainly. Thank you.

HAITHAM EL NAKHAL:

We are not waiting till - we will not be waiting till that we have only two /8. We are talking now for getting approve of this proposal, so IANA will reserve 10 /8 out of its free pool for the last allocation. OK, so if this proposal get approved, so we - IANA will reserve these 10 /8s out of its free pool for the last allocation for each RIR, two /8s for each one, OK.

The second part of the question is why we need such proposal? Am I right?

ROB SEASTROM:

I'm not sure that this proposal does anything, actually.

HAITHAM EL NAKHAL:

OK. So the answer is to be able to plan for projects that could be done in the coming years, and for newcomers in the region as a start till IPv6 become a dominant in Internet traffic, as we don't know anyone could predict that when IPv6 is coming dominant on the Internet.

RANDY BUSH:

Excuse me. Let me rephrase Rob's question - what will you know then that you don't know now?

ROB SEASTROM:

Thank you, Randy.

RANDY BUSH:

What policy will you set then that, if you know it's a good policy, you set it now. Is that fair, Rob?

ROB SEASTROM:

Yeah. Thank you.

IZUMI OKUTANI:

Yes, I get your point, but we don't feel the need to reserve the rest of the remaining IANA pool for the portion. It's only a small portion. We want to keep the existing policy as much as possible until the last minute. However, the concern we have, if we carry out everything under the current policy until the end, we might not be able to address some of the issues we face at the time of exhaustion, like the ones I mentioned, like maybe those without a single IPv4 address cannot connect to the Internet and we probably don't need a whole lot of remaining /8s. I think a single /8 will probably be good enough. That's why we keep it this way.

Does that answer your question?

ROB SEASTROM:

So what you're saying is that your internal assessment of the right time to panic is when you have two /8s left?

IZUMI OKUTANI:

No. Not necessarily. We can get prepared for now and we can be informed of the policy change from now. This can - we can start this from now.

ROB SEASTROM:

Let me say that another way.

There will be a change in your behaviour as an RIR when there are two /8s left? Is a global policy really needed for that? Or can you just decide that - to me, this seems like a decision that you can make internally at 1:30 in the afternoon you're going to go and have lunch if you haven't gotten around to it already.

IZUMI OKUTANI:

I'm not sure if I fully understand.

ROB SEASTROM:

If you have the answer to the rationale behind it.

GEOFF HUSTON:

This is Geoff Huston from APNIC. I believe I can answer Rob Seastrom's problem - I'm not speaking in favour or against the proposal. I trying to address what I believe the question is.

What's going on right now is you've got five RIRs going to IANA at different timings and different intervals getting space. And the scenario that they're thinking about is when you're down to the last, let's say, eight, and it's then getting very, very close to the last few allocations, at this point, for example, RIR number three might only have a quarter of a /8 left free. But its consumption rate is quite slow. Whereas RIR number four might be getting through two /8s every week. And when they go to IANA and get the last two, RIR number three is still sitting with only half a /8 and, if they wanted, if you will, to reserve an equal amount of space for each RIR to do something undefined special. Have I got that right? In the last two /8s, you want to reserve the ability for each RIR to do something different to what they're doing? Yes.

In the model without the 'hand out all the rest', one RIR might not have even two /8s left to do something special so what they're trying to do is say, at that last point, give every RIR the same last one, the same amount, so they can do something special. I'm not saying good or bad.

RANDY BUSH:

Geoff, two points. First of all, Izumi-san's proposal is one /8. We're repeating a mistake, so I don't want to do it.

And the example I might choose is, when it's even closer, there are six left and ARIN gets two or three, right? Everybody gets surprised because all of a sudden somebody who thought they were going to have one more and had planned for it in their strategy and worked out a very good plan, Rob, will find, oops, doesn't work.

And that is, I think, when I characterise the underlying reserve strategy is to remove that surprise and only that surprise.

IZUMI OKUTANI:

Thank you for the clarification. That's exactly the intention, Randy.

ROB SEASTROM:

OK. Well, the last thing, which was the last bullet point on the still-not-retrieved slide, is the concern about RIR shopping. And I guarantee you that that will be your next concern once you have an organisation that has two /8s in it and everybody else is running out. And the older RIRs have a much higher burn rate than the younger RIRs, and I am concerned that there will be a repeat of other natural resource runs by the usual consumers geopolitically, so you need to think about how to protect the remaining space for the use of that RIR's constituency.

IZUMI OKUTANI:

Yes, and, in answer to your point, I totally agree and we were exactly concerned about exactly the same point and that is why, with our proposal, we limit it to a single /8. And with a single /8, there is a possibility that, you know, depending on the timing of the request, that a single RIR could have gotten zero and then, or two - the buffer is between 0 to 2, so, by making it 1, I don't think it will have the actual impact as you - if what you're concerned.

RANDY BUSH:

Rob, can I interrupt a sec. I would like to remove the RIR shopping discussion entirely, because the discussion essentially says that the person who is poor - we should not give them a plate of food, because somebody could steal it. So we should just starve them ourselves.

ROB SEASTROM:

No, that has nothing to do with my point. My point is that, if you know that you are getting something of value, and you want to preserve the use of that something of value for a particular purpose, that you need to have a plan in place to make sure that you keep it.

RANDY BUSH:

The poor person should plan for protecting their plate, yes.

ROB SEASTROM:

So I think the discussion of two versus one, I think that frankly, you're going to have a problem with that either way and that it is something that you should be careful of.

BILL MANNING:

Bill Manning. I think I'm wearing an ARIN hat today. I will be an ARIN person today.

In another context, I don't - let me state, I don't like either of these proposals because of their flatness of their nature.

In a different region with the same type of discussion, there was a suggestion that the IANA, instead of handing out /8s, as we approached the end, we get to the last ten, that they adjust the allocations to the RIRs downward to, like, a /19, and so the RIRs get smaller pools as the IANA pool runs out.

This actually - I think - solves some of the same problems that these proposals are trying to solve without actually having to deal with what appears to be the large /8 problem. You're just dealing with /19s, then it might be a more tractable and graceful slowdown.

If you haven't heard that, I'd be happy to send pointers.

PAUL WILSON:

Paul again here. Just to respond to Bill in case it's not entirely obvious - that type of proposal, to ask IANA to subdivide IPv4 address space in blocks smaller than /8 is a very fundamental change to the method of dealing with the IANA at this stage, operationally, and in a policy sense. There would be global policy implications, there would be implications in terms of the reverse DNS for one thing. We'd be asking IANA to do a substantial amount of extra work they don't do at the moment. And I think it would involve global discussion at a global policy level, which we should consider in terms of how long, how complex and how long that could take.

TAKASHI ARANO:

Takashi Arano from JPNIC. Actually, if I - the flat number is a very good question and, you know, whilst trying to answer. Actually, in my view, my personal view, that this number should not be based on past consumption rate. There should be a future consumption. There may be population or some other factors, but, if it's not clear, just basically one is a very good choice.

That's the reason for that one number.

And why am I saying so is that actually after exhaustion date, IPv4 network will be still big, big portion of the Internet, right? So under one technical findings or fact is that even with just one group of IPv4 address, newcomers cannot start service. Assume that the - you are newcomer and you get just IPv6 address only. And the address you can use is just IPv4 private access or IPv6 global address, right?

And if you can connect your network to the IPv4 Internet, probably you will set up a translator or NAT, right, but you need at least one global IPv4 address.

So, doing nothing implies that we will shut out newcomers in the Africa or Latin America or even in Asia Pacific. This is very important point. That's why we are making this proposal. Thank you.

JORDI PALET:

Jordi Palet. I agree with what just Takashi said but my belief is that it will reach that point. I really think we will not reach that but it will reach that point. This newcomers will be able to get this single IPv4 public address from the upstream provider, OK? Because they will need to buy transit and the transit provider, as part of the service, will be able to deliver this address. So I don't think the newcomer -

RANDY BUSH:

Which they get from where?

JORDI PALET:

Sorry?

RANDY BUSH:

Which they get from where?

JORDI PALET:

Well, the upstream provider, they will have reserves from the pools for probably a couple of years and they will renumber the infrastructure if necessary, they will release some of the IPv4 addresses they are using because they will move to, for example, all the core IPv6. This is going to happen for sure. And then they will free, they will get their own addresses which today are used in infrastructure to be released for the customers.

HAITHAM EL NAKHAL:

Excuse me, I have one question - do you mean by single IP, one address?

JORDI PALET:

One address if necessary.

HAITHAM EL NAKHAL:

Excuse me, excuse me. As a customer, I would like my customer connected from two providers, so I couldn't take a /24.

JORDI PALET:

OK. I am talking in the worst case one address. Probably it will be just a /24.

HAITHAM EL NAKHAL:

A newcomer is an LIR so you would like to give an LIR one single IP address to serve its customers? With single IP address?

JORDI PALET:

Yes, NAT and IPv6, translators and IPv6.

HAITHAM EL NAKHAL:

What if the customers need web hosting, host server? All this is done with NAT?

JORDI PALET:

You can host has many as you want with a single IP address.

RANDY BUSH:

Excuse me. This discussion has happened in many forums. I don't think we're in agreement on it, to be polite, and could we move on? This is not the IPv6 SIG.

JORDI PALET:

OK, in any case, what is important is to understand that we are talking about two years or three years from now and the situation will be much different.

HAITHAM EL NAKHAL:

I couldn't predict the future.

RANDY BUSH:

We'll see. Ray. Who are you?

RAY PLZAK:

Ray Plzak from ARIN. It seems to me in hearing this whole discussion going on, is that this is absolutely the incorrect discussion to be having. The correct discussion to be having is the discussion that was introduced by Izumi in the first part of the talk, which is the much more complicated problem to solve, is the relationship between the RIR and their various customers, because some of their customers are very large. Some of them are very small. Some of them have particular policies that they're operating under. And it seems to me that the time should be spent trying to figure out what you're going to do there, because then you can come up with the determination about what my final end piece should look like and then, in turn, make that as a question under the current policy, global policy, as, "This is what my need is going to be," as opposed to sitting here and saying, "OK, we are going to say that at a certain point, each RIR gets one or two or I don't care what the number is amounts of address space," and then they have to figure out how they're going to whack it up from there.

You should be figuring that out right now. And the discussion should be not what the IANA's give in to the RIRs. It should be what the RIRs are going to do with the address space they have already. So I think the more serious discussion is the topic that Izumi introduced this morning and that that is probably what should be discussed as one or more policy proposals and, at that point, you can figure out what impact this RIR strategy would have on the global pool.

RANDY BUSH:

Would anybody from this region care to speak to the subject?

PHILIP SMITH:

I'm from this region. Philip Smith, Cisco. I live in Australia. Honestly. I do

RANDY BUSH:

No, you don't. You live on airlines.

PHILIP SMITH:

OK. Apart from Singapore Airlines.

OK, my view - I'm quite strongly against this proposal. The best number to end with is zero, for all raft of reasons I've expressed on the APNIC policy SIG mailing list. The biggest issue that I have is that we're going to proposal this, say we do N equals one or N equals two, nobody's figured out with what we're going to do with the N equals one or two bit. We've got random vague Conservative policies. Who defines them? Is it going to be the membership here? Is it going to be the membership of the existing LIRs and their lawyers or whoever else? Is it going to be the people with the biggest chequebooks? All this kind of thing is a really big deal.

And until these two proposals come up with something like this, they are wasting time in this forum.

RANDY BUSH:

I would encourage somebody else from the APNIC region to come to the mic.

IZUMI OKUTANI:

OK, so I have questions to the people on the floor here.

So, with our case, we thought that it would be useful for RIRs or LIRs to have a more predictable, you know, to make the remaining pool of each RIRs - make it more predictable, but you don't think this is, these proposals will help to make it moor predictable and do you think it would be useful to have it more predictable? That's the first question. Second, do you think that this particular proposal would help make it more - does it help in making it more predictable about the remaining RIR? Pool? So I have two questions.

Maybe I wasn't clear enough. Did you get my point, Randy? Maybe you can help me clarify my points?

RANDY BUSH:

I think I could phrase this as a question to the audience.

Which is, let's reduce these proposals to - there will be some very small number that will be handed out at the end, evenly, and evenly because there are 36 factors, you'll never figure them and you'll never get them right, so you just do it even, OK? Or maybe there's a better to say there's a fixed amount you're going to get at the end. And this is to prevent that surprise, OK? And let's presume that Philip's question of what - how that last space is allocated within each region is done by the policy procedures of that region. How do people feel about the basic question of "Is the predictability of everybody's going to get two scoops of rice on their plate at the end and evenly"? Does that appeal to people? Or do people feel this is not workable? OK? Philip? Go for it.

PHILIP SMITH:

I'm quite happy for each region to receive fair scoops. I just don't know what the definition of 'fair' is. Mean, I can't predict the future like Arano-san, but I can go on what the past is and I tend to say, "Well, if APNIC needs so much per week in the past, then that should be the final fair scoop that APNIC gets. "

RANDY BUSH:

So let me challenge you with this question, Philip:

If you think the fairness of the scooping is so important, would you chair a committee to decide what's fair? Or would the - the hypothesis behind my question is that's not really a decidable question.

PHILIP SMITH:

It's not really a decidable question. We could spend the next five or ten years deciding what fair is. The same as the definition - and I saw it in a commentary on one of the mailing list. "What is 'critical'?" that's another rat hole we could end up going round. Everybody's got their own definition.

SAMREEN FATIMA:

Since nobody from subcontinent is here, so I just want to just focus on one point. We all are discussing whether we fairly give scoops to each RIR or not. And people who are opposing this proposal, either two scoops or one scoop on their plate in the end, are forgetting one thing - that people living in Pakistan, people living in India, people living in Bangladesh, with huge number of population and increasing growth rate for Internet traffic, are still not accepting the fact that these pools are going to deplete.

So reserving these pools would at least ensure that they will get whatever their share from each RIR within their region to last for the end. So if we don't evenly reserve pools for each RIR, then we will be at a great loss just because of ignorance. Although APNIC is doing a lot to increase the awareness of IPv4 repletion and moving on with other things, but we will still, living in this region, are not accepting it, are not adapting to it, are not putting in their resources to find ways for smooth transition. If we don't reserve these pools, these regions would suffer because of ignorance.

RANDY BUSH:

Thank you. Ray?

RAY PLZAK:

Ray Plzak, ARIN. Geoff, could you adjust your model so that the end point is not 0, but it's like maybe 5 or 10 and then you would have a different date to set up whenever it was, in 2010? I mean it's the same issue here? It's what point do you decide that things end? This whole policy proposal is about the IANA pool and it's not at 0, but at some other number. And so, from a predictability stand point, you still have the same prediction problem, but you're moving the date a little further to the left. You're still doing the same thing and you still would have the same issue, which is, when you get to this point - if you went to the 0 point, someone would come up and there's no more late and someone who was there just before got their full scoops of rice.

But now, instead of getting there when there's none for me, it's only one for me and the person who was there just before me got more. So it's still the same issue.

RANDY BUSH:

But you can plan you will get that block.

RAY PLZAK:

I can plan I'm going to get 0 too. It's a matter of I'm going to get some or I'm not going to get any, and so it's a matter of what are you planning for? OK? I can plan not to be able to get any just as easily as I can plan for only getting one.

So I think to put some predictability into this, you can't, I guess, put predictability in this because all you're doing is changing what the end point looks like. The end point is still there. The getting there is still the same. It's just a matter of how do you arrive there and I would presume that Geoff could very quickly change it and tell us that, instead of 2010, it's now 2009. I don't know.

IZUMI OKUTANI:

I understand Ray's point in the view that, with any prediction you can't be completely accurate and it can change so maybe it won't work that much. But I think it would at least help reduce buffer to a certain extent. For example, if you don't know if you're going to get zero or two, then there's a difference between the date up to six months, but we can minute mite that buffer at least. That's the idea behind our proposal.

RAY PLZAK:

All I'm saying is I can plan for 0 just as easily as I can 1. I can plan for the day when I go to the well and the well is dry, OK? And, plan how I want to approach that point. And whether that is 1 or 0 or 2 is immaterial, I still have to plan for that point. And so I am the one who has to decide what the buffer is and I have to do that through internal policies about how I distribute that address space to my customers, because I have a variety of customers. And so that is where the real issue is.

RANDY BUSH:

Geoff.

GEOFF HUSTON:

This is Geoff Huston responding to this from APNIC. If you change the question of what the work I've done is really about, then I'm precisely where Ray is.

If what I'm trying to predict with that is the date at which the current allocation policy framework is no longer applicable, then, if you pull ten /8s out at the last and say, "They're reserved to some other policy." Then he's quite right, instead of being May 2010, it's February 2010. It brings that date forward, because the real date that that thing predicts is not exhaustion. What the practical information behind that date is when does the business we currently do in the policy framework we currently operate no longer fly? When does that big bucket in the sky that we allocate this way no longer apply? And Ray is right in that respect, that, from that perspective, from that question in particular, that interpretation, pulling out space for particular purposes that are not consistent with the current allocation framework simply brings the date forward. It doesn't change the fact that that's the date

RANDY BUSH:

Geoff, given there are approximately 40-something /8s left, and Okutani-san's proposal says, "At five left, they'll be evenly distributed, instead of distributed to first come first served," do you estimate that it would change the end date significantly?

GEOFF HUSTON:

Your question and Ray's are actually quite subtly different.

RANDY BUSH:

Not subtly.

GEOFF HUSTON:

There are a few assumptions behind both that are different. And the answer is, at five, when you consider that, in that year, and two years hence, you're talking about... approximately two /8s a month going through, yes? No, it would be one /8 a month, five /8s, it would be about 40 days. A very good number, actually. And 40 nights.

RANDY BUSH:

Oh, it's the nights that get you, man! Especially when you're up awake wondering how to transition to v6. Paul.

PAUL WILSON:

Paul from APNIC. Both Geoff and Ray are acting as though - they seem to be assuming that the date, whenever it is, of exhaustion of the IANA pool results in an identical situation. But, of course, it doesn't. If we allow the IANA pool to exhaust to zero, then we have a certain situation where the RIRs have all used that IANA pool. They only have what is left in their covers.

If we bring that date forward and have a predicted allocation to each RIR that they know is coming, that they can plan ahead for, then the date comes when the IANA pool is exhausted, but the RIRs at that point have got some address space to deal with. So it's a very different situation. It's not just bringing the date forward. It's bringing the date forward and allowing a predictable situation in which certain policies can be implemented.

Philip asked a very good question about what policy would be implemented and Izumi gave a good, I think, summary across a whole range of possibilities there and it represents some good thinking, I think.

So my point would be some of those policies that Izumi mentioned are actually unrealistic and impractical in a situation where we are running towards a final zero exhaustion of IANA space. APNIC could not possibly choose to adopt a very Conservative approach, for instance, to use of the remaining address space. If, by doing that, all we're doing is allowing other RIRs to continue to exhaust the rest of the IANA space at the same rate as they are today, there would be absolutely no point, no benefit. In fact, it would be a great disbenefit if we were to today decide we were going to implement a policy to slow down our consumption of IPv4. It would be a gift to the other RIRs that we would be thankful for, but not a gift to this community.

So I think that there is a big difference between the final end date and the end date that's brought forward and that big difference is our flexibility and our autonomy within this region to decide how that last amount of address space is going to be used. We might decide not to change it at all, in which case the final date is probably largely unchanged. We may decide and we could start talking about that right now. We may decide to do different things with that address space.

So that's the comment I'd like to make, supporting Philip's call for policies but I think while this question is open, we really don't know what policies are in scope and viable and what policies we might as well forget about.

IZUMI OKUTANI:

Thank you. I think that's a very good summary. Thank you.

GAURAB RAJ UPADHAYA:

Thank you. Gaurab. I don't know that this proposal has not been as active on the mailing list but I don't see any reason behind this policy but because it seems like we're planning for eventual death and trying to postpone it, rather than preparing for what's going to happen next. So, you know, I adjusting the pool at the end doesn't make any difference to me. Like Ray says, plan for when the well is going to be dry, don't plan for when the well has one bucket of water left in it and that pretty well sums my view of on. Thanks.

ADARSH SINGH:

Good evening. Adarsh. I'm thinking about all the end-users. I come from ACL, a company which is trying to tie up with a company like Microsoft and Red Hat, and trying to have PC under sub10 K, which is under $200 US. The telecom in India is booming. I mean it's the lowest rate across the world and fibres and metro Internet, WiMAX, are being spread across the country. I really don't know if the two scoops which you're talking of would actually fill up the country like India, where actually the Government of India is trying to target 300 million broadband customers by 2001.

So the good part is that we have two people out here who are actually making certain policies and we are debating on it, why it's flat, why it's not taking in the population or the geopolitical stuff. But somehow, we have to have a policy which is much more in-depth, I feel, and which caters into the requirement of the current growth of a country, economy, region and not in specific on this country. It could be the Latin also that the gentleman over here was talking of. So the policy needs to have more valuables. I'm not trying to be sceptical about it. I'm just trying to say that it needs to have much more maturity and much more depth and it's not a simple straightforward solution that says, "If we add X plus Y and we get Z out of it." And that's it. Thank you.

RANDY BUSH:

May I ask you a question, Sir?

If the discussion of what other model, other than flat, is so complex and so long that we run at of IPv6 space before we finish, are you willing to lose the base of the proposal which is to remove the surprise?

ADARSH SINGH:

No, Randy. I got the point. The point I see is that we've come up with two flat structures. At least we have something out here that we're debating on. I think that's a good point. But the two scoops of rice distributing equally - I don't know how to make sense. Paul was demonstrating about a person with a fat chequebook who might be able to talk to APNIC or IANA. I don't know. It's going to be quite political I feel. I feel that's one perspective. But if we don't have some kind of strong, fundamental thing that says this is the way we actually go about and I completely agree with Gaurab's point also, we see when we are going to hit the wall. It's very clear. A as for the introduction we have out here. Specifically in India, we go with the NAT policy. Mark my words, there could be 500K customers behind three public addresses of NAT. And it works out here.

So I don't know. If you see that IP addresses are going to get exhausted in the next two or three years, we've got to have some kind of policy, maybe that creates pressure on the companies, the ISPs or others to move to IPv6. That's my take on this. Thanks.

RANDY BUSH:

Thank you. Somebody else from the region?

SAMREEN FATIMA:

I agree with Philip that we should have a much broader policy on how to distribute. But the policy step would come next. First, we have to agree, should we reserve IANA's pool or are we not? I would go for the reserve part, because I know that this region would not respond until they are two inches before the wall.

So, at that time, whether we have any policy, whether it's the fat chequebook which decides who's going to have the last free pools, or whether it's the country's growth or the regional growth or whatever it is, we can make a policy then, but at least we should reserve the pool for those of us who - I am sounding a little sceptical, but for those who only respond in panic situation, and we are not yet panicked yet. It's as simple as that. We'll respond when we reach the last five pools and then whether it's the chequebook which decides the policy or whether it's whatever policy region the RIRs make or whether the policy IANA makes, that still has some time, it would buy us time to make those policies, but the region only responds in a panic situation, and the region is still not yet panicked.

RANDY BUSH:

I would say that the region is not homogenous and some parts of the region are obsessive about planning and, in fact, that part of the region has presented this proposal.

Ming-Cheng, I think you were next.

MING-CHENG LIANG:

Hi, this is Ming-Cheng from TWNIC. I think essentially I like this idea proposal by JPNIC, because this is to prepare for the end. But I do have a slight comment, that maybe in, out of five, I would think that the last, the last part, maybe five /8s or six /8s, I was thinking that maybe we need to include IANA as part of this team also, maybe something additional.

But anyway, maybe if it is possible that in the proposal we could have something, what should we do with this /8? Maybe it's a separate RIR policy, could be, but should we have more limited usage or something like that? Maybe make this proposal will be more attractive. Thank you.

IZUMI OKUTANI:

Thank you. And that was the issue I was actually trying to address in my informational presentation and we totally agree that we need to decide on that as well. So thank you for your input.

JONNY MARTIN:

Jonny Martin. I'm against this policy, largely for the same reasons that have come up from others. Just one of the reasons why I'm against it is there is a reason I would support it, that would be for example, if it would be one or five. I don't see that it's achieving what we need to achieve now. It's waiting till the last moment. It hinges on what's deemed as fair but we've established that we're not going to try and determine what is fair today.

SUMON AHMED SABIR:

This is Sumon from Bangladesh. To me, this discussion is very important if you consider that this is not really going to work. If it works well for us, we can adopt it quickly. So is it really very important to discuss what is happening at the consumption of the IPv4 space. It is too early to put this together. We can eventually, when, for some time, we can eventually decide. I don't know if I've read it wrong, but to me this discussion is not very important at this moment.

RANDY BUSH:

I'm not sure I understood that. Were you saying that it is important to agree on the division now and that we can defer the discussion of policy for how the last is handled within the RIR or are you saying we do not need to discuss the division yet?

SUMON AHMED SABIR:

Actually, to my understanding, it is important how we can implement v6 quickly and run it together with v4. If we can run it smoothly, then it's not really important. So the important thing is to figure out how quickly we can implement this and adopt it. So even, if it gives two away, maybe it will linger it for some time more to exhaust the IPv4 addresses.

So the important thing is how quickly we can add up and how quickly we can use v6.

RANDY BUSH:

Personal question. Have you deployed IPv6?

SUMON AHMED SABIR:

Yeah, I am trying actually. We have a deployment from APNIC. We are deploying in our office but we're having some problems deploying to our customers. We are applying for more v4, but we are thinking we can go on v6, why are we trying for v4 any more. I'm not sure if we're right or wrong. That's what I'm trying to get from this forum. I've asked people but didn't really get any clear answer on that.

ADARSH SINGH:

I've just got one question for the - from the two people out there who has proposed the policy. How much time do we have, just in approximate, and how much time would we have after we have come to the figure that we have only five /8s left? And, assuming that we use this policy of the flat structure, we distribute the /8 to all the RIRs and how much time we have to act to convert to IPv6 platform? Do we have that kind of sufficient window space to actually move from IPv4 to IPv6?

RANDY BUSH:

Can I step in a second? I think you have two questions there, which is how - when the last, when we get the last /8, how long before that is consumed? Geoff, how quickly do we consume a /8? Don't bother going to the mic, I'll repeat for you? How fast do we eat /8s under current policy?

GEOFF HUSTON:

We're currently consuming at a rate of one /8 a month.

RANDY BUSH:

In APNIC?

GEOFF HUSTON:

No, in the world.

RANDY BUSH:

In APNIC.

GEOFF HUSTON:

In APNIC, we're using six /8s per year at this point.

RANDY BUSH: