You're viewing an archived page. It is no longer being updated.
RIPE Meeting: |
65 |
Working Group: |
Address Policy |
Status: |
Final |
Chair: Sander Steffann, Gert Doering
Scribe: Denis Walker (RIPE NCC)
Session I - Thursday, 27 September, 09:00 - 10:30
Session II - Thursday, 27 September, 11:00 - 12:30
Gert Doering, APWG co-Chair, opened the session and introduced the audience to the "no more Address Policy Working Group". He reminded the audience that they do have IPv6 addresses and these still need policies to be made. The RIPE 64 APWG draft minutes were approved and the WG chair thanked the scribe. Gert then presented the agenda.
Jan Zorz, go6, asked if the discussion on "2012-04 (IPv4) PI assignments from the last /8" could be moved from the last slot in the first session to the second session. This was to avoid the conflict with the IPv6 WG session running in parallel to the first APWG session. Gert suggested he could move items "H. IPv4 maintenance policy" and "Y. Open Policy Hour" to the first session if Nick had no objections. There were no objections. There were no further changes to the agenda.
- Emilio Madaio, RIPE NCC
The presentation is available at:
https://ripe65.ripe.net/presentations/242-PolicyUpdate-RIPE65.pdf
There were no questions.
- Alex le Heux, RIPE NCC
The presentation is available at:
https://ripe65.ripe.net/presentations/244-alex_le_heux_-_apwg.pdf
Gert asked when the IANA pool of 16-bit ASNs would run out.
Alex replied that he didn't have that number available immediately.
Leo Vegoda, ICANN, said that there were four blocks available for allocation to the RIRs which should last about 12-18 months roughly.
Alex le Heux explained that even though they are into the last /8, much of the allocation and assignment policies still apply. Alex asked if we should try to clean up old policies.
Wilfried Woeber, Vienna University, said they should try to clean up the mess of old policies. Some of them contradict each other. Many of the texts have inconsistencies. These policies are difficult to understand for anyone who may not have participated in the background discussion on them over the last 15 years.
Gert thanked Wilfried for his very clear message.
Rob Blokzijl, RIPE Chair, said he also saw a need for a clean up. Rob started to work on this with the RIPE NCC, but the run out of IPv4 put a lot of workload on the RIPE NCC staff. Maybe it is time to continue with this work. Most policies relate to allocation. As there are no addresses
left, all they need now is the registration part. He said he wanted to aim for one clear document explaining the history and what is still relevant today.
Nina Hjorth Bargisen, TDC A/S, said she supported the idea of cleaning up old assignment policies and thinks it should be an urgent and prioritised action on the working group. The current policies, which are at the moment still in effect, are very strict and can put long delays
in the operations of making assignments to customers.
Gert explained that this is why Tore Anderson put forward the 2012-06 policy proposal, which reverses the “run-out fairly” policy for the assignments. The run-out policy has no use anymore. He suggested that maybe they could fast track this proposal if people agree with it without much discussion. The working group should voice their opinion. Gert thanked Nina for making this point clearly.
Hans Petter Holen, Visma, commented that since everyone seems to agree that something needs to be done, they need a proposal for this on the table and a quick decision.
Gert reminded the group that there is the policy proposal 2012-06.
Alex concluded his presentation. There were no more questions.
- Gert Doering, APWG co-Chair
Gert reminded the audience of the IPv4 maintenance policy presented at RIPE 63. Since then not much has happened because everybody was too busy with the IPv4 run-out. The community dealt with a number of IPv4 related policy proposals on the table. From a purely process perspective, the idea was to wait for the dust to settle and then do the complete overhaul of the policy framework with the maintenance document.
No questions were asked.
- Sandra Brown, IPv4 Market Group
The presentation is available at:
https://ripe65.ripe.net/presentations/237-RIPE-2012-02_and_03Preso.pdf
Gert reminded the audience that no decisions are made at meetings. At the RIPE meetings the APWG co-chairs hear you, listen to you and it helps reach some conclusion on which way to go with a policy proposal, but everything needs to be on the mailing list. The mailing list is open and consensus is based on what happens on the list.
Sandra reminded the audience that APNIC and ARIN already have proposals, transfers should be accurately reflected in Registration Database, and extending transfer period from three to 24 months would be inline with thinking across three regions.
Sandra concluded stating that she looked forward to have the proposal in place by January.
Gert remarked that the APWG co-Chairs needed feedback from WG to move this quickly because silence is not full consensus.
Remco Van Mook, Equinix, encouraged people to support it.
Ingrid Wijte, RIPE NCC, asked where the three months implementation plan came from and that no impact analysis had been done yet to see how much work is involved.
Gert confirmed that it looked a bit optimistic, but could be done if there is enough support and everyone moves quickly.
Ingrid responded that the RIPE NCC always try to do it ASAP, but it depends on different factors.
Gert recognised that there is lots of work to do behind the scenes and that takes time of course.
Nina, TDC, stated that this looked like a new allocation policy and that maybe it should be considered as part of the old policy cleanup. She supported the idea of transfers.
Gert admitted that he saw support and no strong opposition and invited everyone to continue on the mailing list. The next version will include comments already made on mailing list that raised no objections.
Remotely, Tore Anderson, Redpill Linpro, commented that this policy only covers allocations, at least for now; there are no policy proposals on assignments.
There were no more questions.
- Gert Doering, APWG co-Chair
Gert continued with some more change to the agenda, back to topic of the IPv6 PA/PI unification policy. The simplification was getting complicated. No policy proposal was ready yet. He will try to take it to the mailing list in coming weeks or months.
Daniel Stolpe, Resilans, asked if there were any specs on what the goals of this policy were.
Gert replied that the goal was to have a simple and straightforward policy where all address blocks are the same, size depends on your needs, but the rules are the same.
Daniel Stolpe completely agreed. He commented that today's rules are strange and don't match reality.
Gert warned that taking care of all the side effects of simplifying the rules was slowing down progress, but expect to have a very straightforward IPv6 policy in the end.
There were no more questions.
Gert continued chairing the meeting and mentioned the IPv6 PI growth after the approval of the proposal 2011-02, “Removal of multihomed requirement for IPv6 PI”. He illustrated that no significant change when this policy was approved. He showed graphs with steady growth.
Erik Bais (A2B Internet), as author of the policy, wanted to comment that this was what he was expecting.
No questions were asked.
Temporary IP assignments - Nick Hilliard, INEX
The presentation is available at:
https://ripe65.ripe.net/presentations/249-inex-ripe-apwg-amsterdam-pi-tempassignments-2012-09-27.pdf
Nick presented the need to update the temporary assignment policy. The time scale in the policy created dissatisfaction in the community: temporary address space can be received only one week before the planned use. People need more time in advance to set up their network so this should be changed. He wanted to collect feedback from the audience on the three options he had in mind:
Remco Van Mook, Equinix, suggested not aiming for the perfect solution but simply deciding a new fixed time scale and call it done.
Tore Anderson, Redpill Linpro, commented remotely that he disapproved of the over-engineered booking system; he added that if someone can't use space early he or she can't actually debogonize it.
Nick confirmed.
Wilfried Woeber, University of Vienna, recommended the simple option with the least risk of routing problems, accordingly to the level of the demand it looked like a non-issue so he suggested not making it complicated. He said that a generous, well-defined fixed timescale was sufficient to solve the problems.
Nick replied that the IPRAs did not see not a huge uptake of requests. He agreed with Wilfried on the simple and generous time period. However, he added that the reality now is different since no addresses are available anymore. The demand may grow in the future. He also stated that the feedback seems to be in favour of generous time scale and a simple policy change.
Gert asked the RIPE NCC to give some input.
Alex Le Heux, RIPE NCC, commented that an impact analysis was not done yet. He could confirm that the booking system would be hard to implement and complicated. Furthermore, in terms of the advance planning, there is the potential issue that new address space may leak into routing tables. Removing the timescale can create confusion about the correct timings without guidelines for IPRAs. He concluded that changing timescales may lead to disagreement over reasonable “time”, but is the best option.
Gert concluded that from the room it looked like most people wanted the option to change the timescale.
Sander Steffann, AWPG co-Chair, added that we should be generous with the fixed period to make most people happy.
There were no more questions.
IPv4 sub allocations - Alexey Ivanov, Leader Telecom
The presentation is available at:
https://ripe65.ripe.net/presentations/254-Presentation_ripe.pdf
Alexey presented the idea to remove from the suballocation policies the limit of /20 once every 12 months. This would provide a flexible alternative to transfers.
Remco Van Mook, Equinix, commented that all this is covered and can be done within the transfer policy. He did not see any need for another policy
Alexey responded that transfer needs work by the RIPE NCC, this option just needed users to make changes in RIPE Database.
Remco replied that IPv4 holders still have to demonstrate need. However, he was ready to read and give feedback to a policy proposal.
Sander invited Alexey to take it to the mailing list for further discussion.
There were no more questions.
- Tore Anderson, Redpill Linpro (via Skype)
The presentation is available at:
https://ripe65.ripe.net/presentations/231-20120926-RIPE65-Policy_Proposal_2012-06_Revert_Run_Out_Fairly_After_IPv4_Depletion.pdf
Remco Van Mook, Equinix, said he supported the proposal.
Gert stated that they had support on the mailing list and no objections. The proposal was still in the Discussion Phase. When this was over, very likely they would decide to go forward with this policy.
- Nick Hilliard, INEX
The presentation is available at:
https://ripe65.ripe.net/presentations/270-inex-ripe-apwg-amsterdam-pi-lastslash8-2012-09-27.pdf
Wilfried Woeber, University of Vienna, said he agreed with the aim of this policy: money buys you a /22 by becoming an LIR, even if you only needed a /24. As policy is worded now, one could apply for PI and then give it away. Wilfried said he would not support it without some additions.
Nick said that the transfer policies do not allow for transfer of directly assigned PI addresses but that may change in the future. They need to try to prevent abuse of policies.
Wilfried added that he would like the community to think about the bigger picture and form a consistent overall system.
Erik Bais, A2B Internet, commented that the current policies may be not fair, but this policy prevented new companies from getting address space in the future and he did not support it.
Nick stated that he did not agree that LIRs are more important than other users. There were degrees of unfairness in a time of starvation.
Erik responded that if an End User wants address space they could become an LIR.
Nick commented that becoming an LIR provides you a /22. An organisation can keep a small piece and sell all the rest. It was better to give them a /24 if that is all they needed. He added that maybe there was a middle ground.
Erik added that he did not disagree that this was an ideal solution. However, by alllowing PI in the last /8 they will speed up the use of that last /8.
Nick pointed out that the run-out is going to happen anyway, they can't change that. He did not expect the speed of consumption to matter.
James Blessing, Limelight, remotely asked if imposing a time limit before allowing a transfer would be a solution.
Nick replied that he was not of the opinion that this would change the policy very much and really fix the problems we have.
Remco responded to James' question pointing out that transfer was only allowed for PA at the moment. It was better to not engineer for future possible problems.
Remco said he also welcomed the idea of seeking fairness but any such proposal will never get consensus. There will be many people saying yes and no. He wanted to see address space used as effectively as possible, regardless of speed. Giving out small pieces rather than aggregated blocks would not help anyone.
Gert commented that LIRs were treated differently as they are assumed to have many users behind them.
Nick agreed that there is no such thing as fairness. Consensus does not require unanimity. He said he was still optimistic over the possibility of consensus.
Hans Petter Holen, Visma, thought that they were looking at the symptom and not at the root cause. He asked whether people wanted to pay for the running of the RIPE NCC or just pay for a small piece of address space. He added that he did not see how the concept of PA/PI would be sustained in the future. Trading smaller and smaller addresses would cause routing issues. They needed a completely new framework.
Nick stated that it was hard to go to PI holders and put their costs up so much by becoming an LIR. He agreed with Hans about the current system being messy.
Yakovenko Volodymyr, Google, requested the data information about how much reclaimed address space was actually returned since last year and how many LIR assignments have been made from the last /8. He said that he also considered the information on how many PA requests were received versus how many attempts to get PI space useful. He also suggested the idea of considering a new policy of address distribution from the pool of reclaimed space and not from the last /8.
Nick responded that the suggestion was a possibility but was also messy. He asked what would happen when real last /8 was depleted, adding that as per the requests of data information he found those interesting numbers however they did not change the philosophy behind the policy
proposal.
Yakovenko disagreed adding that this should be a data driven discussion.
Alex le Heux, RIPE NCC, announced that last year the reclaimed address totaled about 800,000. The RIPE NCC gave out about 2.5 million PI addresses a year.
Alexey Ivanov, Leader Telecom, commented that the last /8 was the last space available and if they only gave it to LIRs it might last several years. But if they gave out PI it would run out quickly.
Nick disagreed and said that there would be a market so addresses will still be available but at a cost.
Alexey responded that money was a good way of regulation: depending on the availability there will be a price for address space. He said that in his opinion, giving out PI would cause more fraud cases.
Nick commented on the inevitability of the run-out: when the Titanic sank people ran to the lifeboats.
Ivan Beveridge, Betfair Ltd, asked if there was a regulatory concern over PI, if anyone could become an LIR then there are no restrictions.
Nick replied that the RIPE NCC has a good relationship with governments and regulators. The RIPE NCC has always implemented fair and open policies. The last /8 policy might attract attention from regulators especially if a disgruntled end user makes a complaint.
Ivan responded that an end user could get address space from an existing LIR or even from other RIRs. He added that RIPE was not restricting it to a closed club and anyone could become a member.
Hans Petter Holen, Visma, wondered how they were dealing with PI space in v6.
Nick answered that one can just apply for it and get it.
Tore Anderson, Redpill Linpro, remotely asked how many new users have signed up as LIR and if this rate has increased.
Alex reported that about 20 new members signed up since the RIPE NCC reached the last /8. He added that the process of becoming an LIR can take one week or more so some maybe signed up before the reach of the last /8.
Elliot Lear, Cisco, remotely commented that in the absence of lawyers in the room, he suggested referring this matter to the council and did not see any reason why one would consider the policy based on the regulatory threat.
Nick invited the audience to look at Alex's presentation during the morning on the number related to the final /8 implementation.
Lorenzo Colitti, Google, agreed with Ivan and added that an existing channel to get address space was open to anyone and has not changed.
Nick responded that it is an issue of perspective and there are organisations that have business needs for a small number of PI address space.
Juergen Kammer, eurodata GmbH & Co. KG, expressed his concerns with the concept of fairness when dealing with old PI holders and wannabe PI holders. He said that another problem he had with fairness was the price of PI space: he wondered why PI holders should pay EUR 50 and not ¼ of what a LIR pays since they get ¼ of a /22, that is a /24.
Nick responded that the concept of fairness is difficult to define. As per the cost of the PI space he considered it a policy issue for the RIPE NCC to deal with. He concluded that he was of the opinion that the EUR 50 per PI space was, in his opinion, just a fraction of the whole cost for Registration Service and administration of the registry necessary to implement for example policies like 2007-01.
Juergen emphasised also the importance of getting more severe restrictions to get PI space in order to reduce the cases of abuses.
Nick responded that it is impossible to create a set of rules that cannot be abused.
Remco, speaking as treasurer of the RIPE NCC Executive Board, said he preferred not to put a price tag on IPv4 addresses.
Gert wrapped up and stated that the problem was clear but this is not the proposal the community wants. He invited the WG to elaborate some more input and then the proposer will be able to come up with another version of the proposal that is more acceptable.
- Milton Mueller (via Skype)
Gert presented the forthcoming new text of the proposal and added that not much feedback was expressed on the mailing list but there was a clear request to record rejected transfers in anonymised form only. Then he passed the presentation to Milton.
Hans Petter Holen, Visma, expressed his support. He also added that they should look at the same openness and transparency on pricing as well.
Milton responded that maybe pricing would be the next step. This transfer listing may end up being done by intermediaries and not the RIPE NCC.
Gert commented that the pricing information could be difficult to retrieve if we imagine that the RIPE NCC has to ask to publish it.
Gert invited taking it to the mailing list, including anonymising rejected transfers
There was no additional business.
Gert closed the session by thanking the audience for attending and inviting them to the next RIPE 66 meeting in Dublin.