A comment on 2008-03 & 2007-09
Afternoon, First off let me say that I am broadly in agreement with 2008-03, but I wished to make clear my thoughts on the incompatibility of the two proposals in the subject. I see 2008-03 as an exercise in fairness, a way to evenly distribute the last few /8s, rather than reaching a point where, particularly, AfriNIC and LACNIC need more addresses and find there to be none. However it strikes me that this policy is completely incompatible with 2007-09. If both policies were introduced then I can easily envisage a scenario where a bigger RIR uses up its /8, then starts to nibble away at the remaining addresses of those who will be slower to allocate their space, ie AfriNIC and LACNIC, thus defeating the purpose of fairness that I see inherent in 2008-03. The worse case scenario here, for the less developed RIRs at least, is that they may see very little of that last /8. I realise that many people have different views of the impetus behind 2008-03, but I can only view it through my eyes, so apologies if anyone feels I'm attributing motive where there is none. Regards, Brian.
I see 2008-03 as an exercise in fairness, a way to evenly distribute the last few /8s, rather than reaching a point where, particularly, AfriNIC and LACNIC need more addresses and find there to be none.
more than that, it is an exercise in planning. an rir can count on having one last /8 instead of hitting the wall in surprise when their sibling got there ten minutes prior.
However it strikes me that this policy is completely incompatible with 2007-09.
this is not an accident. the author of -9 is in extreme opposition to -3 and has crafted -9 to very intentionally nullify and circumvent -3. it is notable that -3 got massive support in the arin meeting, and is generally supported in the other regions though not yet passed, and -9 has been pretty much rejected worldwide. randy
Randy Bush wrote:
I see 2008-03 as an exercise in fairness, a way to evenly distribute the last few /8s, rather than reaching a point where, particularly, AfriNIC and LACNIC need more addresses and find there to be none.
more than that, it is an exercise in planning. an rir can count on having one last /8 instead of hitting the wall in surprise when their sibling got there ten minutes prior.
A lesson in not writing emails when one is in a hurry somewhere. The comment on it being equally an exercise in planning was lost in the editing process. I fully agree with you.
However it strikes me that this policy is completely incompatible with 2007-09.
this is not an accident. the author of -9 is in extreme opposition to -3 and has crafted -9 to very intentionally nullify and circumvent -3.
it is notable that -3 got massive support in the arin meeting, and is generally supported in the other regions though not yet passed, and -9 has been pretty much rejected worldwide.
Then I was veering into stating the obvious, but it was not something I had seen stated fully before. Thanks for the additional background. Brian.
Brian, On May 30, 2008, at 5:23 AM, Brian Nisbet wrote:
If both policies were introduced then I can easily envisage a scenario where a bigger RIR uses up its /8, then starts to nibble away at the remaining addresses of those who will be slower to allocate their space, ie AfriNIC and LACNIC, thus defeating the purpose of fairness that I see inherent in 2008-03. The worse case scenario here, for the less developed RIRs at least, is that they may see very little of that last /8.
Suppose we fast forward to ~2011 and you've just been rejected by RIPE- NCC because they have no more address space to hand out whereas AfriNIC and LACNIC both have (at least) a full /8. I'm curious: what do you think is going to happen? Thanks, -drc
At 02:45 p.m. 30/05/2008, David Conrad wrote:
Brian,
On May 30, 2008, at 5:23 AM, Brian Nisbet wrote:
If both policies were introduced then I can easily envisage a scenario where a bigger RIR uses up its /8, then starts to nibble away at the remaining addresses of those who will be slower to allocate their space, ie AfriNIC and LACNIC, thus defeating the purpose of fairness that I see inherent in 2008-03. The worse case scenario here, for the less developed RIRs at least, is that they may see very little of that last /8.
Suppose we fast forward to ~2011 and you've just been rejected by RIPE- NCC because they have no more address space to hand out whereas AfriNIC and LACNIC both have (at least) a full /8.
I'm curious: what do you think is going to happen?
David: I am curious about other situation. Suppose that IANA allocates the last 2 /8s to the RIR A and one day later IANA receives a request from the RIR B that is running out of IPv4 addresses while the RIR A has (at least) 2 /8s. What do you think is going to happen? Raúl
Thanks, -drc
-- No virus found in this incoming message. Checked by AVG. Version: 7.5.524 / Virus Database: 269.24.4/1474 - Release Date: 30/05/2008 07:44 a.m.
On May 30, 2008, at 11:11 AM, Raul Echeberria wrote:
Suppose we fast forward to ~2011 and you've just been rejected by RIPE- NCC because they have no more address space to hand out whereas AfriNIC and LACNIC both have (at least) a full /8.
I'm curious: what do you think is going to happen?
David:
I am curious about other situation. Suppose that IANA allocates the last 2 /8s to the RIR A and one day later IANA receives a request from the RIR B that is running out of IPv4 addresses while the RIR A has (at least) 2 /8s.
What do you think is going to happen?
My understanding of the current policy regime is that IANA only allocates unused addresses. With the allocation of the last /8, IANA would be out of the IPv4 business. As such, it would be up to the RIRs involved to negotiate some "fair" solution (if any). Regards, -drc
My comment was in the sense of comparing the scenario with the policy approved vs. the scenario without this policy. Of course, it is not the end of the discussion, if the policy is approved, many things will have to continue being discussed. BTW, the policy passed in LACNIC region in this meeting and now it will go through the last call for commetns and ratification process. Raúl At 03:19 p.m. 30/05/2008, David Conrad wrote:
On May 30, 2008, at 11:11 AM, Raul Echeberria wrote:
Suppose we fast forward to ~2011 and you've just been rejected by RIPE- NCC because they have no more address space to hand out whereas AfriNIC and LACNIC both have (at least) a full /8.
I'm curious: what do you think is going to happen?
David:
I am curious about other situation. Suppose that IANA allocates the last 2 /8s to the RIR A and one day later IANA receives a request from the RIR B that is running out of IPv4 addresses while the RIR A has (at least) 2 /8s.
What do you think is going to happen?
My understanding of the current policy regime is that IANA only allocates unused addresses. With the allocation of the last /8, IANA would be out of the IPv4 business. As such, it would be up to the RIRs involved to negotiate some "fair" solution (if any).
Regards, -drc
-- No virus found in this incoming message. Checked by AVG. Version: 7.5.524 / Virus Database: 269.24.4/1474 - Release Date: 30/05/2008 07:44 a.m.
Raul, On May 30, 2008, at 12:25 PM, Raul Echeberria wrote:
My comment was in the sense of comparing the scenario with the policy approved vs. the scenario without this policy.
The situations aren't analogous. There is only one IANA from which the RIRs obtain addresses. There are 5 RIRs and there are many organizations which obtain addresses from more than one RIR. For example, if a service provider in Europe were to establish a legal presence in (say) Namibia and request address space, what restrictions would be placed upon the announcement of that address space and how would AfriNIC enforce those restrictions (or even determine they were being violated)? The point is that I view Tony's proposal as an attempt to pragmatically deal with the reality that businesses are likely going to do what they can to ensure they meet customer requirements, regardless of the arbitrary geographical monopolies put in place when there was more address space than people could count. Questions of "fairness" get extremely complex when you get into resource scarcity (e.g., fair to whom? Will it be fair that a charity NGO based in Geneva will be unable to get any address space whereas DeBeers will be able to get as much as they desire?) and I personally think it naive to assume that the folks with money aren't going to be able to get what they want, regardless of "set asides"... Regards, -drc
David. You didn't answer my question. My point is that not having a policy of this kind would be very problematic at the time of allocating the last part of the pool. It is not necessarily a question of fairness but certainty in how much space the RIR will have since a certain moment to the end of IPv4 space. Is this the right policy, I don't know, the community has to say that. In ARIN and LACNIC the policy has received enough support and it is in the end of the process. I would not like to make more opinions about the policy itself until the process is finished in LACNIC region, but let me say that the rest of your arguments are based in assumptions that are not proven and are questionable. Raúl At 05:43 p.m. 30/05/2008, David Conrad wrote:
Raul,
On May 30, 2008, at 12:25 PM, Raul Echeberria wrote:
My comment was in the sense of comparing the scenario with the policy approved vs. the scenario without this policy.
The situations aren't analogous. There is only one IANA from which the RIRs obtain addresses. There are 5 RIRs and there are many organizations which obtain addresses from more than one RIR. For example, if a service provider in Europe were to establish a legal presence in (say) Namibia and request address space, what restrictions would be placed upon the announcement of that address space and how would AfriNIC enforce those restrictions (or even determine they were being violated)?
The point is that I view Tony's proposal as an attempt to pragmatically deal with the reality that businesses are likely going to do what they can to ensure they meet customer requirements, regardless of the arbitrary geographical monopolies put in place when there was more address space than people could count. Questions of "fairness" get extremely complex when you get into resource scarcity (e.g., fair to whom? Will it be fair that a charity NGO based in Geneva will be unable to get any address space whereas DeBeers will be able to get as much as they desire?) and I personally think it naive to assume that the folks with money aren't going to be able to get what they want, regardless of "set asides"...
Regards, -drc
-- No virus found in this incoming message. Checked by AVG. Version: 7.5.524 / Virus Database: 269.24.4/1474 - Release Date: 30/05/2008 07:44 a.m.
Raul, On May 30, 2008, at 1:54 PM, Raul Echeberria wrote:
You didn't answer my question.
Apologies, I misunderstood.
My point is that not having a policy of this kind would be very problematic at the time of allocating the last part of the pool. It is not necessarily a question of fairness but certainty in how much space the RIR will have since a certain moment to the end of IPv4 space.
I was mostly commenting on assertions on Tony's policy (2007-9, in RIPEland), not on 2008-3. With respect to 2008-3, I understand the sympathize with the desire to ensure that RIRs don't get surprised at the end of the free pool. As I've stated in the past, I don't have a problem with the policy, particularly with N=1 as it is now. However, with that said, I still believe the statement: "Concerns could be raised that explicitly allowing regional policies will encourage RIR shopping. However, this should not happen if the requirements within each region is adequately reflected in each RIR's policy through PDP. RIR may also chose to add criteria to prevent LIRs from other regions submitting such requests." ignores the likely desperation people will face when we reach IPv4 runout, particularly given the mechanisms the RIRs have used in the past to encourage policy conformance (i.e., potentially refusing to allocate additional address space if policies aren't conformed to) would no longer be applicable. Regards, -drc
I am curious about other situation. Suppose that IANA allocates the last 2 /8s to the RIR A and one day later IANA receives a request from the RIR B that is running out of IPv4 addresses while the RIR A has (at least) 2 /8s.
I don't think that IANA would ever do this because it would be an incredibly stupid thing to do. Since I have never heard IANA being accused of gross incompetence before, I have to believe that when they receive that request from RIR A, they will ask all the other RIRs about how many addresses they need before allocating anything. In fact, I expect IANA to do this well before the last 2 /8 blocks. Then, after consulting all RIRs, I expect IANA to publish their intentions and ask for comments before allocating any blocks. This is just good business practice and I don't think that the RIRs need to write policies which tell IANA to do this. --Michael Dillon
Michael, On Jun 1, 2008, at 2:59 AM, <michael.dillon@bt.com> <michael.dillon@bt.com
wrote: I don't think that IANA would ever do this because it would be an incredibly stupid thing to do. Since I have never heard IANA being accused of gross incompetence before,
Well, IANA staff get accused of this less often these days (:-)).
I have to believe that when they receive that request from RIR A, they will ask all the other RIRs about how many addresses they need before allocating anything.
Current policy does not dictate IANA do this. As the v4 free pool is depleted (particularly as the free pool approaches 10 /8s), I suspect (hope?) there will be greater communication amongst the RIRs.
In fact, I expect IANA to do this well before the last 2 /8 blocks. Then, after consulting all RIRs, I expect IANA to publish their intentions and ask for comments before allocating any blocks.
This would be a significant change to existing policy and as such, is not something IANA could implement unilaterally. Whether there is sufficient time left for this to be defined as a "global policy" might be an interesting question.
This is just good business practice and I don't think that the RIRs need to write policies which tell IANA to do this.
I suspect you misunderstand the relationship between the RIRs and IANA. Regards, -drc
I suspect you misunderstand the relationship between the RIRs and IANA.
Actually, I thought I was talking about the relationship between ICANN and IANA. --Michael Dillon
David Conrad wrote:
Suppose we fast forward to ~2011 and you've just been rejected by RIPE-NCC because they have no more address space to hand out whereas AfriNIC and LACNIC both have (at least) a full /8.
I'm curious: what do you think is going to happen?
As you note, the concept of "fairness" is a rather difficult notion. Neither 2008-03 nor 2007-09 are going to stop the inevitable squabbling that's going to happen: "It's not fair - they have a /8 and we don't have any. Waah!", or "I'm not going to share my /8 with you. No, you CAN'T HAVE IT! Waah! I'm going to tell on you!! Daaaaaddddyyyyyyy!" Of course, I don't want to trivialise the problem - it is rather a serious one. But fair is generally perceived as fair for someone when they benefit more from the arrangement than anyone else. Otherwise "fair" is not remotely fair. Perhaps "fair" just means what we want it to mean at a particular point in time, and from the point of view of deciding on a end-game v4 address pool distribution strategy, what we define as "fair" now may not be perceived as fair when one's local RIR runs out of addresses. I have a preference for 2007-09 over 2008-03 - at least the former has the advantage that the RIR exhaustion dates are more-or-less simultaneous worldwide, which is equally (in)equitable to all. RIR exhaustion will happen quickly and it will cause a shock, and then we'll just have to move on. As regards 2008-03, I'm not going to object to it, but I suspect it is more likely to cause protracted disgruntlement in the future until every last block of address space has been allocated from all RIRs, and this could be a matter of a year or two. But there you go: the proposal is popular in all other regions and each RIR community is fully within its remit to hoist itself on its own petard. As a side issue, it will be interesting to see in a couple of years time whether the loudest complainers (whether end-user or LIR) in the v4 exhausted world will be those who have made the most preparations for depletion. Nick
Nick Hilliard wrote:
David Conrad wrote:
Suppose we fast forward to ~2011 and you've just been rejected by RIPE-NCC because they have no more address space to hand out whereas AfriNIC and LACNIC both have (at least) a full /8.
I'm curious: what do you think is going to happen?
As you note, the concept of "fairness" is a rather difficult notion. Neither 2008-03 nor 2007-09 are going to stop the inevitable squabbling that's going to happen: "It's not fair - they have a /8 and we don't have any. Waah!", or "I'm not going to share my /8 with you. No, you CAN'T HAVE IT! Waah! I'm going to tell on you!! Daaaaaddddyyyyyyy!"
The more general observation is that there is no objective concept of "fairness" in such a situation. So as we try and figure out what changes (if any) should be made to the current distribution framework within the time available the challenge is that we are never going to be able to make the increasingly finite pool of remaining IPv4 addresses comfortably encompass the continuing sequence of needs that are expressed in address allocation requests. So if the aim of such tweaks is to maximise "fairness" then I for one get lost pretty quickly in understanding precisely what that means. And if the challenge is to make the finite become infinite, then we are not exactly making good use of what time remains. It's not that I'm trying to belittle the various arguments here, but what appears to me is that there is a certain shuffling of perceived future cost burden here - it appears that many folk see no advantage in early adoption of IPv6 and indeed perceive it as a penalty and a cost, and are therefore wanting to secure their own future source of IPv4 resources with the attendant consequence of forcing others into a position of necessity to confront IPv6 deployment sooner rather than later. This is being played out within each RIR (large vs small allocation debates, for example) and across the RIRs with policy debates such as these. However, its not clear (to me) that there is any overwhemlingly "right" answer here, nor is it clear (to me) that with more time to cogitate and debate the issue that we'll come up with any such solution. What is clearer (to me) is that what we are lacking here is a general sense of confidence that we can make this transition operate efficiently, effectively and safely, and part of the fuel for this debate over the last /8 may be interpreted as a perceived reluctance to just get on with what needs to be done in terms of network and service engineering. My suspicion is that if we were more confident that we understand what transition really meant, in terms of engineering, products, services, infrastructure, business, competitive positioning, lines of supply, etc, etc, then this entire policy discussion over what remains in the IPv4 pool would perhaps be of lesser importance in the grander scheme of things Internet. Geoff (speaking entirely for myself, naturally!)
David Conrad wrote:
Brian,
On May 30, 2008, at 5:23 AM, Brian Nisbet wrote:
If both policies were introduced then I can easily envisage a scenario where a bigger RIR uses up its /8, then starts to nibble away at the remaining addresses of those who will be slower to allocate their space, ie AfriNIC and LACNIC, thus defeating the purpose of fairness that I see inherent in 2008-03. The worse case scenario here, for the less developed RIRs at least, is that they may see very little of that last /8.
Suppose we fast forward to ~2011 and you've just been rejected by RIPE-NCC because they have no more address space to hand out whereas AfriNIC and LACNIC both have (at least) a full /8.
I'm curious: what do you think is going to happen?
I suspect that any answer I would give to such a question would be very non-standard due to the research and education space in which I work and given that our core network has been dual stacked for some years and we're doing everything we can to educate our clients and extend the v6 rollout as quickly as possible. So the scenario above is very unlikely to come to pass. In a different organisation the answer may well be "get addresses from anywhere we can" leading to situations under potential market conditions or an attempt to make oneself a legitimate customer of a different RIR, but both are ultimately exercises in holding back the tide for a little while. 2007-09 is much the same thing, in my eyes, while 2008-03 allows for a little more vision of when things are going to happen and possibly even a reduction of the inevitable panic. And considering some of the issues over PI space contracts and ERX addresses and the like that's going on right now, I dread to think what's going to happen if the space for which an organisation or LIR is contracting actually comes from that RIR over there via an RIR which is, in fact, a LIR. Brian.
participants (7)
-
Brian Nisbet
-
David Conrad
-
Geoff Huston
-
michael.dillon@bt.com
-
Nick Hilliard
-
Randy Bush
-
Raul Echeberria