IPv6 Working Group Minutes - RIPE 92
Date: Thursday, 21 May 2026 (UTC+0100)
Chairs: Christian Seitz, Wolfgang Tremmel, Raymond Jetten
Scribe: Florian Obser
Status: Draft
View the stenography transcript
1. Welcome
The presentation is available at: https://ripe92.ripe.net/programme/meeting-plan/sessions/94/LAZ8VR/
Co-chair Christian Seitz opened the session and introduced the agenda. The minutes from RIPE 91 were approved.
The chairs celebrated IPv6's 30th birthday and pointed out that IPv4 is only 16 years older.
2. The year of IPv6-only desktop
Ondřej Caletka, RIPE NCC
The presentation is available at:
https://ripe92.ripe.net/programme/meeting-plan/sessions/94/PBHCLK/
Ondřej gave a historical overview of transitioning to IPv6 on the RIPE meeting network. He noted that since RIPE 85 in 2022, the main conference network is running IPv6-mostly and clients could opt out of requesting an IPv4 lease since then. He showed statistics that the number of clients doing so has increased at every RIPE meeting. At RIPE 92, 85% of clients opted out of IPv4.
Ondřej pointed out that we need proof that IPv6 is working and dual-stack is not providing this proof since IPv4 is always there as a fallback.
Ondřej finished the talk by giving an overview of client support for IPv6-mostly: Android, iOS & macOS fully support it, Linux will get support with NetworkManager 1.58 and Windows will get it in version 11. He pointed at the next talk for more details. He mentioned that we are missing IoT devices and smart home devices.
Jen Linkova, Google, asked attendees to share their experience with IPv6-mostly deployments. She also asked whether the increase in clients opting out of IPv4 could be explained by MAC address randomisation, resulting in devices being counted multiple times.
Ondřej responded that he could not determine this from the data because the MAC addresses were randomised. However, he did not believe this was the primary explanation, noting that MAC address randomisation had been introduced before the widespread adoption of IPv6-mostly deployments.
Benedikt Stockebrand, Stepladder IT Training+Consulting GmbH, commented that IPv6 reduces complexity by eliminating the need for multiple layers of NAT. He argued that the community has moved beyond the point where IPv6 itself is a significant source of operational problems.
Ulrich Wisser, ICANN, observed that Ondřej had said dual-stack deployments can mask IPv6 issues because IPv4 continues to function when problems occur. He noted that this could leave the impression that IPv6 remains problematic. Ondřej responded that improvements will only come through broader deployment and use of IPv6.
An attendee pointed out that the next step is to remove the IPv4 stack from operating systems. Ondřej replied that there are experiments in Linux, but it is hard to remove all IPv4 dependencies.
Yannis Nikolopoulos, ORIZON TELECOM SA, remarked online that dismissing dual-stack as a transition technology in 2026 is unfair since it was the only thing we had for 20 years.
Robert Scheck, unaffiliated, mentioned online that NetworkManager 1.58 will be merged into Fedora shortly after RIPE 92.
Gert Doering, SpaceNet AG, thanked the RIPE NCC team for their continued efforts to deploy IPv6 on the RIPE Meeting network, and their willingness to proceed despite operational issues and their work to resolve problems as they arise.
3. IPv6-only Network Design
Wilhelm Boeddinghaus, CEO Route 128 GmbH
The presentation is available at:
https://ripe92.ripe.net/programme/meeting-plan/sessions/94/XJWGCT/
Wilhelm opened his presentation with an overview of the three main components of an enterprise network. He explained that the core network is predominantly dual-stack, while the data centre is also largely dual-stack, typically using separate network interfaces for each address family. In contrast, the office WLAN is operated in dual-stack mode only where necessary and generally uses a single network interface.
Wilhelm then described a typical office desktop environment consisting of Windows PCs and laptops connected to a large, flat Layer 2 network. He outlined some of the security challenges associated with this type of deployment.
Wilhelm explained IPv6-mostly and WinCLAT, currently available in closed beta. He mentioned that WinCLAT chose the current beta of choosing an IPv4 address out of the 192.0.0.0/29 Service Continuity Prefix and embedding this IPv4 address directly into the CLAT IPv6 address. This reduces the number of clients to 8.
He then showed the currently needed protocols to run an IPv4 LAN, and pointed out that running a dual-stack network for the office network adds more protocols.
Wilhelm proposed a simplification by using IPv6-mostly with micro segments using layer 3 switches. He explained that with this scheme every client runs in an isolated network segment. He noted that this approach allows a number of protocols to be removed, simplifying the overall network design.
He also argued that IPv6-mostly deployments can improve security, as malware is no longer able to perform reconnaissance using Layer 2 mechanisms.
Wilhelm concluded the presentation by reviewing the issues identified in RFC 9898. He noted that the proposed network design addresses most of these challenges.
Jen Linkova, Google, pointed out that one has to be careful when moving to layer 3 networks, as some protocols might want to talk to their neighbours. She mentioned IoT as an example. Wilhelm responded that one has to test the deployment and that the talk is not about IoT at all, since IoT is very special.
Marc van der Wal, Afnic, asked for clarification why WinCLAT only supports 8 clients on the network. Jen Linkova explained that WinCLAT maps the IPv4 /29 directly into an IPv6 prefix instead of using a random 64-bit interface identifier. Wilhelm mentioned that Microsoft promised a fix in the next beta release.
4. IPv6 Routing Loops Update
Maynard Koch, TU Dresden
The presentation is available at:
https://ripe92.ripe.net/programme/meeting-plan/sessions/94/LXPTTU/
Maynard gave an update on IPv6 routing loops and amplification. The amount of routing loops went down from 156 million /48 prefixes to 146 million /48 prefixes between RIPE 90 and RIPE 91 and then went up to 149 million prefixes for RIPE 92. He noted that 10.9 million prefixes could be used for amplification attacks due to router bugs.
He continued to explain routing loops and noted that solving routing loops will also solve the issue of amplification.
He finished the presentation by asking people to check their networks for routing loops and fix them.
Éric Vyncke, Cisco, asked which types of routers were affected by the bug that made them susceptible to amplification attacks.
Maynard responded that the issue primarily affected older Juniper and Cisco routers. He said that the root cause had been traced to a Broadcom chipset and that Broadcom had classified it as a product limitation rather than a software defect and therefore did not intend to address it.
5. The Killer App for IPv6
Benedikt Stockebrand, Stepladder IT Training+Consulting GmbH
The presentation is available at:
https://ripe92.ripe.net/programme/meeting-plan/sessions/94/THKMSP/
Benedikt opened the presentation by arguing that IPv4 is not the “killer application” for IPv6, while noting that IPv6 is nevertheless necessary to sustain the continued operation of IPv4.
He explained that the focus of his presentation was Layer 2 networking. Reflecting on his experience as a trainer, he noted that Layer 2 switching had traditionally been faster than Layer 3 routing. However, he said that the introduction of Layer 3 switching had largely eliminated this distinction.
Benedikt argued that many operational problems can be avoided by moving to Layer 3 switching. He suggested that tagged VLANs could be replaced by technologies such as VXLAN or, potentially, Segment Routing. He also noted that, when combined with micro-segmentation, these approaches can reduce the need for multicast routing. In addition, he argued that dynamic routing protocols offer better convergence times and scalability than Spanning Tree Protocol (STP) variants.
He concluded by encouraging attendees to read RFC 1925, “The Twelve Networking Truths”.
There were no questions.
6. Thanks, Wrap Up and Rate the Talks
The presentation is available at:
https://ripe92.ripe.net/programme/meeting-plan/sessions/94/C3GWMU/
Raymond Jetten closed the session by reminding people to rate the talks.