2010-05 New Policy Proposal (Global Policy for IPv4 Allocation by the IANA post exhaustion)
Dear Colleagues, A new Global Policy Proposal has been made and is now available to the RIPE Community for discussion. You can find the full proposal at: http://ripe.net/ripe/policies/proposals/2010-05.html We encourage you to review this proposal and send your comments to <address-policy-wg@ripe.net> before 20 September 2010. Regards Emilio Madaio Policy Development Officer RIPE NCC
* http://ripe.net/ripe/policies/proposals/2010-05.html
The reclamation pool will initially contain any fragments that may be left over in IANA inventory.
I assume these fragments are the so-called «VARIOUS» blocks? These, that is: <http://bgp.potaroo.net/ipv4-stats/prefixes_various_pool.txt>?
The Reclamation Pool will be divided on CIDR boundaries and distributed evenly to all eligible RIRs. [...] [...] an RIR will become eligible to request address space from the IANA Reclamation Pool [...]
These two statements appear to me to contradict each other. The first seems to imply a push-based model, where the Reclamation Pool (RP) is divided in five equally large chunks and dealt out to the RIRs at the same time, in a manner very similar to the procedure in 2008-03. The second seems to imply a pull model, much like how the IANA today allocates /8s to the RIRs upon request. (If so, how much can a RIR allocate at a time? 18 months worth of consumption, like today?) Which is it? And if it is the latter, I fail to see how it will be distributed «evenly». It's unrealistic to assume that all RIRs will deplete at the same time. Isn't it then more likely that the first RIR(s) to deplete their inventory (and thus activate the RP) will burn through all of the RP long before the last RIR depletes its inventory and gets eligible for making use of it? Finally there's proposal 2010-02 ... If that one is accepted as currently prosposed, RIPE won't be eligible to receive anything from the RP until there's 16384 LIRs in the region who have all received an IPv6 allocation and a /22 from the last IPv4 /8. At that point, I'm not sure if anybody really cares about what happens to the remaining IPv4 space, especially as none of the 16384 LIRs that have already received their /22 from the last /8 can come back for more anyway. Should the space set aside for the implementation of 2010-02 (and similar policies in the other regions) be disregarded when determining when a RIR has exhausted its inventory? That is, a timeline like the following: 1) Receive last /8 from the IANA - set aside for 2010-02 2) Allocate/assign remaining inventory to LIRs according to ripe-492 3) Receive 1/5 (or: continuously allocate until no longer possible) from the IANA RP. Allocate/assign to LIRs according to ripe-492 4) Implement 2010-02 (using the /8 set aside in step #1) Best regards, -- Tore Anderson Redpill Linpro AS - http://www.redpill-linpro.com/ Tel: +47 21 54 41 27
On Sat, Aug 28, 2010 at 11:12, Tore Anderson <tore.anderson@redpill-linpro.com> wrote:
* http://ripe.net/ripe/policies/proposals/2010-05.html
The reclamation pool will initially contain any fragments that may be left over in IANA inventory.
I assume these fragments are the so-called «VARIOUS» blocks? These, that is: <http://bgp.potaroo.net/ipv4-stats/prefixes_various_pool.txt>?
The Reclamation Pool will be divided on CIDR boundaries and distributed evenly to all eligible RIRs. [...] [...] an RIR will become eligible to request address space from the IANA Reclamation Pool [...]
These two statements appear to me to contradict each other. The first seems to imply a push-based model, where the Reclamation Pool (RP) is divided in five equally large chunks and dealt out to the RIRs at the same time, in a manner very similar to the procedure in 2008-03.
The second seems to imply a pull model, much like how the IANA today allocates /8s to the RIRs upon request. (If so, how much can a RIR allocate at a time? 18 months worth of consumption, like today?)
Which is it?
It is a pull model with equal distribution. The key term in the first statement is "*eligible* RIRs." If three RIRs have exhausted their IPv4 address space, they all make that announcement and the subsequent request to the IANA. They then all receive one third of the Reclamation Pool (divided on CIDR boundaries).
And if it is the latter, I fail to see how it will be distributed «evenly». It's unrealistic to assume that all RIRs will deplete at the same time. Isn't it then more likely that the first RIR(s) to deplete their inventory (and thus activate the RP) will burn through all of the RP long before the last RIR depletes its inventory and gets eligible for making use of it?
The intent is to distribute the space evenly amongst all RIRs with need, hence the eligibility requirement. Consider a scenario where three RIRs have exhausted their IPv4 address space and the IANA has nothing to replenish them with. They all declare exhaustion and make their request to IANA. Then someone returns IPv4 space to the IANA. That space would be carved into three blocks and distributed evenly to all three exhausted RIRs. Without this policy, the IANA would have no way to distribute that space at all.
Finally there's proposal 2010-02 ... If that one is accepted as currently prosposed, RIPE won't be eligible to receive anything from the RP until there's 16384 LIRs in the region who have all received an IPv6 allocation and a /22 from the last IPv4 /8. At that point, I'm not sure if anybody really cares about what happens to the remaining IPv4 space, especially as none of the 16384 LIRs that have already received their /22 from the last /8 can come back for more anyway.
Should the space set aside for the implementation of 2010-02 (and similar policies in the other regions) be disregarded when determining when a RIR has exhausted its inventory? That is, a timeline like the following:
1) Receive last /8 from the IANA - set aside for 2010-02 2) Allocate/assign remaining inventory to LIRs according to ripe-492 3) Receive 1/5 (or: continuously allocate until no longer possible) from the IANA RP. Allocate/assign to LIRs according to ripe-492 4) Implement 2010-02 (using the /8 set aside in step #1)
The authors of this proposal (I included) will take this suggestion into consideration. Our largest concern is that this may create a loophole by which an RIR could stockpile IPv4 addresses to the disadvantage of the other regions. Cheers, ~Chris
Best regards, -- Tore Anderson Redpill Linpro AS - http://www.redpill-linpro.com/ Tel: +47 21 54 41 27
-- @ChrisGrundemann weblog.chrisgrundemann.com www.burningwiththebush.com www.coisoc.org
Hi Chris, * Chris Grundemann
It is a pull model with equal distribution. The key term in the first statement is "*eligible* RIRs." If three RIRs have exhausted their IPv4 address space, they all make that announcement and the subsequent request to the IANA. They then all receive one third of the Reclamation Pool (divided on CIDR boundaries).
I see. So, what exactly prevents something like this from happening: Day 1: 0 RIRs eligible, 125.4M addresses in RP Day 2: RIR «Red» depletes its inventory, now only RIR eligible for RP Day 3: RIR «Red» requests and receives its 125.4M/1 share of the RP Day 4: RIR «Blue» depletes its inventory, but can't receive anything from the now-empty RP. Repeat for remaining three RIRs. Best regards, -- Tore Anderson Redpill Linpro AS - http://www.redpill-linpro.com/ Tel: +47 21 54 41 27
On Mon, Aug 30, 2010 at 12:49, Tore Anderson <tore.anderson@redpill-linpro.com> wrote:
Hi Chris,
* Chris Grundemann
It is a pull model with equal distribution. The key term in the first statement is "*eligible* RIRs." If three RIRs have exhausted their IPv4 address space, they all make that announcement and the subsequent request to the IANA. They then all receive one third of the Reclamation Pool (divided on CIDR boundaries).
I see. So, what exactly prevents something like this from happening:
Day 1: 0 RIRs eligible, 125.4M addresses in RP Day 2: RIR «Red» depletes its inventory, now only RIR eligible for RP Day 3: RIR «Red» requests and receives its 125.4M/1 share of the RP Day 4: RIR «Blue» depletes its inventory, but can't receive anything from the now-empty RP. Repeat for remaining three RIRs.
Thanks Tore, A similar concern was raised by folks in the APNIC region and we (the group of authors) are currently discussing some possible changes or additional text to prevent this scenario. We were originally operating under the impression that eligible RIRs would already be in line waiting for space before it was returned. Please let me know if you would like to join the authors group, we would appreciate your ideas on how to make the proposed policy better. ~Chris
Best regards, -- Tore Anderson Redpill Linpro AS - http://www.redpill-linpro.com/ Tel: +47 21 54 41 27
-- @ChrisGrundemann weblog.chrisgrundemann.com www.burningwiththebush.com www.coisoc.org
Hi again, * Chris Grundemann
A similar concern was raised by folks in the APNIC region and we (the group of authors) are currently discussing some possible changes or additional text to prevent this scenario. We were originally operating under the impression that eligible RIRs would already be in line waiting for space before it was returned.
The proposal says the following: «The reclamation pool will initially contain any fragments that may be left over in IANA inventory». You didn't comment on it in your first reply, but if these fragments indeed are the unused space in the legacy /8s (I don't know what else they could be), the initial contents of the RP would be some 125.4M addresses, according to Geoff Huston's IPv4 Address Report (http://ipv4.potaroo.net/). That's a sizable chunk...
Please let me know if you would like to join the authors group, we would appreciate your ideas on how to make the proposed policy better.
I don't really have any suggestions for alternative text for you, but I'd be happy to give you my feedback on any draft updates you might have. Best regards, -- Tore Anderson Redpill Linpro AS - http://www.redpill-linpro.com/ Tel: +47 21 54 41 27
On Mon, Aug 30, 2010 at 14:51, Tore Anderson <tore.anderson@redpill-linpro.com> wrote:
The proposal says the following: «The reclamation pool will initially contain any fragments that may be left over in IANA inventory».
You didn't comment on it in your first reply, but if these fragments indeed are the unused space in the legacy /8s (I don't know what else they could be), the initial contents of the RP would be some 125.4M addresses, according to Geoff Huston's IPv4 Address Report (http://ipv4.potaroo.net/). That's a sizable chunk...
It certainly is. When this topic came up in our initial conversations, we asked an IANA representative about fragments and were told that IANA did not currently have any. My understanding is that this may be due to a deal that was struck between the IANA and the NRO - but I do not have any authoritative information on that at this time. Geoff Huston's report states that: "At the time IANA reaches the last 5 /8s (the "IANA Exhaustion time" as defined by current address allocation policies), these unassigned addresses in the legacy /8s are then distributed evenly to the RIRs." I am not sure where this information came from though, since this is not spelled out in the "Allocation of the Remaining IPv4 Address Space" policy (http://www.icann.org/en/general/allocation-remaining-ipv4-space.htm) as far as I can discern. And of course the IANA report (http://www.iana.org/assignments/ipv4-address-space/ipv4-address-space.xml) shows all of the "various/legacy" /8s as being administered by individual RIRs, since it only has a /8 granularity. So there is some confusion (on my part at least) as to where these fragments currently reside and who has responsibility for them. Perhaps if someone from IANA monitors this list they can chime in? If not, I and the other authors will certainly be following up to iron these fairly significant details out. I apologize for not having a more complete answer.
I don't really have any suggestions for alternative text for you, but I'd be happy to give you my feedback on any draft updates you might have.
Thanks! ~Chris
Best regards, -- Tore Anderson Redpill Linpro AS - http://www.redpill-linpro.com/ Tel: +47 21 54 41 27
-- @ChrisGrundemann weblog.chrisgrundemann.com www.burningwiththebush.com www.coisoc.org
Hi, On 30 Aug 2010, at 2:22, Chris Grundemann wrote: [...]
When this topic came up in our initial conversations, we asked an IANA representative about fragments and were told that IANA did not currently have any. My understanding is that this may be due to a deal that was struck between the IANA and the NRO - but I do not have any authoritative information on that at this time.
Yes, the RIRs agreed a split for the old "Various Registries" space between themselves. They wrote to us about it in 2008 and a copy of the message and the breakdown is published on the ICANN web site: http://www.icann.org/correspondence/wilson-to-conrad-28jan08-en.pdf
Geoff Huston's report states that: "At the time IANA reaches the last 5 /8s (the "IANA Exhaustion time" as defined by current address allocation policies), these unassigned addresses in the legacy /8s are then distributed evenly to the RIRs." I am not sure where this information came from though, since this is not spelled out in the "Allocation of the Remaining IPv4 Address Space" policy (http://www.icann.org/en/general/allocation-remaining-ipv4-space.htm) as far as I can discern.
This is presumably an address management practice agreed by the NRO rather than a matter of policy. Regards, Leo Vegoda
On Mon, Aug 30, 2010 at 15:32, Leo Vegoda <leo.vegoda@icann.org> wrote:
Hi,
On 30 Aug 2010, at 2:22, Chris Grundemann wrote:
[...]
When this topic came up in our initial conversations, we asked an IANA representative about fragments and were told that IANA did not currently have any. My understanding is that this may be due to a deal that was struck between the IANA and the NRO - but I do not have any authoritative information on that at this time.
Yes, the RIRs agreed a split for the old "Various Registries" space between themselves. They wrote to us about it in 2008 and a copy of the message and the breakdown is published on the ICANN web site:
http://www.icann.org/correspondence/wilson-to-conrad-28jan08-en.pdf
Thank you Leo! That was the missing link. I take this to mean that the authors original assumption is correct and that as things sit today, IANA will have no IPv4 addresses remaining immediately following IPv4 exhaustion. The Reclamation Pool will therefor only contain addresses returned to the IANA.
Geoff Huston's report states that: "At the time IANA reaches the last 5 /8s (the "IANA Exhaustion time" as defined by current address allocation policies), these unassigned addresses in the legacy /8s are then distributed evenly to the RIRs." I am not sure where this information came from though, since this is not spelled out in the "Allocation of the Remaining IPv4 Address Space" policy (http://www.icann.org/en/general/allocation-remaining-ipv4-space.htm) as far as I can discern.
This is presumably an address management practice agreed by the NRO rather than a matter of policy.
Understood - and thanks again. ~Chris
Regards,
Leo Vegoda
On 30 Aug 2010, at 2:42, Chris Grundemann wrote: [...]
Yes, the RIRs agreed a split for the old "Various Registries" space between themselves. They wrote to us about it in 2008 and a copy of the message and the breakdown is published on the ICANN web site:
http://www.icann.org/correspondence/wilson-to-conrad-28jan08-en.pdf
Thank you Leo! That was the missing link.
I take this to mean that the authors original assumption is correct and that as things sit today, IANA will have no IPv4 addresses remaining immediately following IPv4 exhaustion. The Reclamation Pool will therefor only contain addresses returned to the IANA.
Yes, as things stand now, that's right. Regards, Leo
On 30 Aug 2010, at 2:57, Leo Vegoda wrote: [...]
I take this to mean that the authors original assumption is correct and that as things sit today, IANA will have no IPv4 addresses remaining immediately following IPv4 exhaustion. The Reclamation Pool will therefor only contain addresses returned to the IANA.
Yes, as things stand now, that's right.
I should probably add a clarification to my previous note. Once those five /8s have been allocated to RIRs we won't have any unallocated unicast /8s. But we will still be managing the multicast space, and various IETF assigned blocks for registries like the IANA IPv4 Special Purpose Address Registry: http://www.iana.org/assignments/iana-ipv4-special-registry/iana-ipv4-special... Regards, Leo Vegoda
Leo Vegoda had a good post on this same topic over on the APNIC list that may answer your question: On Mon 8/16/2010 10:54 AM, Leo Vegoda wrote:
Hi Philip,
On 16 Aug 2010, at 4:37, Philip Smith wrote:
[...]
5.1 Reclamation Pool
Upon adoption of this IPv4 address policy by the ICANN Board of Directors, the IANA shall establish a Reclamation Pool to be utilized post RIR IPv4 exhaustion as defined in Section 4. The reclamation pool will initially contain any fragments that may be left over in IANA inventory. I understood that IANA was exhausting its entire pool. Or is exhaustion really just complete /8s? It would be helpful if someone from IANA could clarify as I was under the impression that remaining fragments would be distributed as well, certainly before IANA declared that the cupboard was bare. The remaining fragments are not insignificant. I can't speak for the authors but I can describe what we "have in stock" and what the ratified policies allows us to do. The only IPv4 address blocks we have with an UNALLOCATED status are whole /8s. Further, the current policy we implement only allows us to allocate /8s:
" * The IANA will allocate IPv4 address space to the RIRs in /8 units." -- http://www.icann.org/en/general/allocation-IPv4-rirs.html
Similarly, the policy for allocating the last blocks required that they be allocated in /8s, too:
"IANA will automatically allocate the reserved IPv4 allocation units to each RIR (one /8 to each one)" -- http://www.icann.org/en/general/allocation-remaining-ipv4-space.htm
My personal interpretation of the text above was the the "Reclamation Pool" would only be a potential pool unless and until one of the RIRs found that they no longer needed the IPv4 address space they were recovering and could return it to IANA for re-distribution. If I've misunderstood the authors' intentions I am sure that they will correct me.
Kind regards,
Leo Vegoda * sig-policy: APNIC SIG on resource management policy * _______________________________________________ sig-policy mailing list sig-policy@lists.apnic.net http://mailman.apnic.net/mailman/listinfo/sig-policy
On Mon 8/30/2010 2:22 PM, Chris Grundemann wrote:
On Mon, Aug 30, 2010 at 14:51, Tore Anderson <tore.anderson@redpill-linpro.com> wrote:
The proposal says the following: «The reclamation pool will initially contain any fragments that may be left over in IANA inventory».
You didn't comment on it in your first reply, but if these fragments indeed are the unused space in the legacy /8s (I don't know what else they could be), the initial contents of the RP would be some 125.4M addresses, according to Geoff Huston's IPv4 Address Report (http://ipv4.potaroo.net/). That's a sizable chunk... It certainly is.
When this topic came up in our initial conversations, we asked an IANA representative about fragments and were told that IANA did not currently have any. My understanding is that this may be due to a deal that was struck between the IANA and the NRO - but I do not have any authoritative information on that at this time.
Geoff Huston's report states that: "At the time IANA reaches the last 5 /8s (the "IANA Exhaustion time" as defined by current address allocation policies), these unassigned addresses in the legacy /8s are then distributed evenly to the RIRs." I am not sure where this information came from though, since this is not spelled out in the "Allocation of the Remaining IPv4 Address Space" policy (http://www.icann.org/en/general/allocation-remaining-ipv4-space.htm) as far as I can discern.
And of course the IANA report (http://www.iana.org/assignments/ipv4-address-space/ipv4-address-space.xml) shows all of the "various/legacy" /8s as being administered by individual RIRs, since it only has a /8 granularity.
So there is some confusion (on my part at least) as to where these fragments currently reside and who has responsibility for them. Perhaps if someone from IANA monitors this list they can chime in? If not, I and the other authors will certainly be following up to iron these fairly significant details out. I apologize for not having a more complete answer.
Someone else can likely provide more detail, but my understanding is that the various /16s in legacy swamp space (the "various/legacy" listings there) are all allocated to (or administered by) individual RIRs, and they're just listed that way because the IANA list doesn't show anything more granular than a /8. So, as far as I can tell, the reclamation pool would initially be empty, unless/until someone returned something smaller than a /8 to IANA. -Scott
participants (6)
-
Chris Grundemann -
Emilio Madaio -
grousseau＠afone.com -
Leo Vegoda -
Scott Leibrand -
Tore Anderson