IPv6 Working Group Minutes RIPE 74

Session 1 - Wednesday, 10 May, 14:00 - 15:30

WG co-chairs: Jen Linkova, Benedikt Stockebrand, Raymond Jetten
Scribe: Alun Davies

B. Do They Follow (the Specs) - A Look at IPv6 Address Selection from the Lab- Enno Rey, ERNW

The presentation is available at:
https://ripe74.ripe.net/presentations/108-ERNW_RIPE74_IPv6_AddressSelection.pdf

Enno Rey reminded everyone that more details on the presentation can be found in a recently published paper on the same topic.

Gert Doering raised a question about selection mechanisms, asking whether Enno is aware of any work at IETF to look at why selection fails when a given path is broken.

Enno suggested that Jen Linkova would be in a better position to respond, but she preferred to continue the discussion offline.

Enno followed up by suggesting that Gert might look at discussion on the mailing list on applying happy eyeballs to IPv6 only scenarios. The suggestion was that Gert might find the scenario he was describing there, but added that he could follow up on this with Gert later on.

Benedikt Stockebrand, in response to Gert, pointed out that there are certain complications that arise when dealing with multiple source addresses. He also invited Gert to contact him offline for further discussion. He mentioned that whilst this solves certain problems, the kinds of problems Gert had raised remain unsolved.

Marcus Keane, Microsoft, pointed out that there's work going on, in the interior, looking at provisioning for domains that have multiple prefixes on the same segment.

Jen Linkova raised two questions. She asked whether Enno had looked back to happy eyeballs for addresses in determining that the machines were making consistent address choices.

Enno responded that they hadn't.

Jen asked whether Enno had looked at cases where two routers advertised two different prefixes and host users.

Enno said that they had not.

Jen commented that she was keen for them to look into this further.

Enno continued, pointing out that rule 5.5 is particularly interesting, then went into more detail about scenarios potentially brought about by 5.5.

Jen added a few further comments on the matter before thanking Enno for his presentation.

C. Corporate IPv6 Number Plan - Friso Feenstra, Rabobank

The presentation is available at:
https://ripe74.ripe.net/presentations/4-Rabobank-corporate-IPv6-numberplan.pdf

There were no questions.

E. Panel Discussion: IPv6 in Enterprise - Panel moderator: Raymond Jetten, Elisa Oyj

(Panel members: Marcus Keane, Microsoft; Enno Rey, ERNW GmbH; Benedikt Stockebrand, Stepladder IT Training + Consulting GmbH; Sander Steffann, SJM Steffann; Friso Feenstra, Rabobank)

Wilhelm Boeddinghaus, iubari GmbH, commented on incentives for enterprises going to IPv6. He asked why they couldn't have IPv6-only in the client networks, running IPv4 as a service over it, adding that they could do this in an enterprise environment and get rid of dual-stack.

Marcus Keane answered, pointing out that, unlike with Android, where there's only a single platform to deal with, enterprise networks have many platforms. From an enterprise point-of-view, there's a lot of stuff that can break in moving to IPv6. VPN support for IPv6 is still lacking, for instance.

Marco Diaz asked a question (via chat monitor) about whether Rabobank will put pressure on Dutch ISPs to enable IPv6 for their users? Friso answered that they will not.

Nathalie Trenaman, RIPE NCC, asked what the most expensive part of IPv6 deployment was for the members of the panel. Friso and Benedikt both answered that it was training, with Friso going into some details on the matter. Marcus Keane added that man hours involved in deployment was a big expense.

Yan Filyurin, Bloomberg, asked a question aimed at consultants on the issue of whether anyone on the panel had discovered new ways of designing network in enterprise due to the options available with IPv6? Benedikt Stockebrand answered that they had, particularly in relation to using smaller subnets. Enno said that they had not identified any new approaches to network design.

Blake Willis, Zayo, commented on an IPv6 rollout plan aimed at alleviating issues with IPv4 address conflicts. There was no response to this from the panel as time was up.

IPv6 Working Group RIPE 74 - Session 2

11 May 2012, 16:00 - 17:30
WG co-chairs: Jen Linkova, Benedikt Stockebrand, Raymond Jetten
Scribe: Mirjam Kühne

B. No v4 Switch Anymore @ Cloudflare -Martin Levy, Cloudflare

The presentation is available at:
https://ripe74.ripe.net/presentations/141-RIPE74_Budapest_Martin_J_Levy_No_v4_Switch_Anymore_at_Cloudflare.pdf

Matthew Moyle-Croft, Amazon Web Services, asked if it would be better to keep the button and find out how why people were switching it off because it might be useful information to have.

Martin agreed and said that they have been collecting this kind of data at Cloudflare and want to investigate it more before removing the switch all together.

Lee Howard, Retevia, said that his guess why people were switching IPv6 off is that they the see something strange on the Internet and try to troubleshoot by fiddling with various buttons. Regarding the Happy Eyeballs issue. Martin mentioned in this presentation, the IETF v6ops WG has adopted a draft on this.
Martin said he was pleased to hear that.

C. Lightning Talk: Active Proposals from BCOP - Jan Zorz, ISOC

The presentation is available at:
https://ripe74.ripe.net/presentations/132-Jan_Zorz-IPv6-prefix-delegations-BCOP-v-2.pdf

Nathalie asked if the BCOP Task Force was planning to publish this as a RIPE Document so it gets a document number.

Jan confirmed that this is the plan.

Ondřej Caletka suggested to continue with more BCP documents, for instance, to advise operators on issues such as “How do I address my LAN in my corporate environment?” and “How do I identify using MAC addresses?” etc,.

Jan responded that, yes, the BCOP TF is planning to continue this kind of work. They gather input in the various RIPE Working Groups and then bring it back to the BCOP Task Force.

Benedikt suggested to use the terminology “longer” and “shorter” prefixes rather than “less” and ”more”.

Jan agreed and the authors will change this in the document.

Jen suggested to issue a WG last call as a next step before this document would become a RIPE Document.

Jordi Palet-Martinez, Consulintel, asked people to send comments as soon as possible (this week and next). Jordi said he would present this at the LACNIC meeting and would like to have an almost final document at that point.

D. NAT64 Experiments - Jan Zorz, ISOC

The presentation is available at:
https://ripe74.ripe.net/presentations/133-Jan_Zorz-NAT64-Check-v3.4.pdf

Ondřej Caletka said that this was an awesome tool and that he was looking for such a tool for a long time. He suggested some improvements and setting up the DNS64 to force people to use it.

Jan replied that he thought that was a bad idea.

Julian Seifert, DNCC, agreed that this was an impressive tool. He said he would further test it and send feedback to Jan and Sander. He added that they were deploying v6-only services more and that this was an interesting tool from a service point of view.

Wil van Gulik thanked Jan and Sander for this tool and offered to further advocate it at other NOG meetings.

Martin Levy suggested to extend the tool to other MTU sizes (other than 1280). He further suggested to use the Alexa-1million list and test them all.

Fredy Künzler, Init7, agreed that this was an awesome tool and asked the audience for a large round of applause.

Magnus Frühling, Freifunk Frankfurt am Main e. V, asked if Jan and Sander saw any special machines where the MTU problem came up.

Jordi was asked to do some beta-testing of the tool and responded that he indeed found some MTU problems.

Jen seconded the suggestion to get the entire Alexa-1million list tested. Jen also commented on some of the examples used in the slides and that they might be confusing to the users. She was also wondering if Jan and Sander checked how many of the broken things are broken because of DNSSEC.

Jan and Sander said that there is many more that could be done with this tool, but that they need more help and that they are looking for volunteers.

E. Using 464XLAT Residential Networks - Jordi Palet Martinez, Consulintel

The presentation is available at:
https://ripe74.ripe.net/presentations/151-ripe-74-ipv6-464xlat-residential-v2.pdf

Lee Howard asked if Jordi or anyone else knew of any modern comparison of various transition technologies (e.g. a table that shows which applications work or break with which transition mechanisms). Most studies are pretty old.

Nobody in the room was aware of such research.

Randy Bush, IIJ, commented that for the residential customers, XLAT and DS-Lite were winning the award for requiring both CGN in the core and forklifting the CPE. He added that it would be great to fix the implementations for Android.

Jen thanked everyone for their participation and closed the meeting.