ipv6-wg

Download

ipv6-wg@ripe.net

November 2022

2 participants
2 discussions

Minutes from RIPE 85 IPv6 WG
by Raymond Jetten 17 Nov '22

17 Nov '22

Dear list members, here are the draft minutes from RIPE 85 IPv6 wg. Please share your comments or remarks to the list so they can be adjusted. The minutes will otherwise be published in two weeks IPv6 Working Group Minutes - RIPE 85 Wednesday, 26 Oct 2022, 12:00-13:00 (UTC+2) WG co-Chairs: Benedikt Stockebrand, Raymond Jetten Scribe: Oleg Muravskiy Status: Draft Welcome, Etiquette, Approving Minutes, Co-Chair Selection IPv6 Working Group Co-Chairs Raymond reminded participants about the Code of Conduct. The RIPE 84 minutes were approved. Benedict announced that Raymond's term as co-chair had ended, but that he was happy to continue in that role. There were no further nominations or objections, so Raymond was re-appointed as co-chair. Path Tracing Pablo Camarillo, Cisco Systems This presentation is available at: https://ripe85.ripe.net/wp-content/uploads/presentations/76-20221026-RIPE85… Pablo described the recent implementation of an addition to the IPv6 packet header extension that records the path of packets in ECMP networks and presented the use cases that were enabled by this feature. Leandro Bertholdo, University of Twente, asked how this addition was different from the IPv4 registration option. Pablo replied that the main difference in this approach was that they recorded the information at the line speed of the basic packet forwarding pipeline, and used only 3 bytes for each node. Leandro asked the presenter if they recorded an ID of the router that the packet passed. Pablo replied that they recorded a sequence of interface ID. Leandro asked if it was for the input interface or the output. Pablo replied that they recorded an ID of the output interface at each one of the routers Alexander Azimov, Yandex, asked if this would work in the MPLS core. Pablo said that this was possible, but that for the moment it was only implemented for IPv6. Open Source NAT64 Implementations Nico Schottelius, ungleich This presentation is available at: https://ripe85.ripe.net/wp-content/uploads/presentations/74-ripe85-open-sou… Nico showed a comparison of different open source implementations of NAT64, and asked the audience to provide more feedback and point out implementations they missed. Jan Žorž, 6connect, asked Nico if they had tried a VPP implementation as it was very fast. Nico said that he hadn't but that it was nice to know about it. Urban Suhadolnik, University of Ljubljana, asked if the presenter knew of any CLAT implementation, possibly for Windows, as the user-space implementation of CLAT was missing. Nico said that he couldn't answer the question as he was not familiar with this the topic. Deploying IPv6-Mostly Access Networks Ondřej Caletka, RIPE NCC This presentation is available at: https://ripe85.ripe.net/wp-content/uploads/presentations/9-RIPE85-Deploying… Ondřej presented an "IPv6-mostly" network that was implemented at this RIPE Meeting, which combined IPv6 functionality only for devices that supported it, and traditional dual-stack for other devices, all in the same network. Leonard Wagner, IN-Berlin e.V, asked since which Android PREF64, DHCP option 108 and CLAT were supported. Ondřej replied that version 11 and above were supported by the vendor, but that he was not sure about older versions. Michael Richardson, one of the authors of the IPv4-only option, noted that when he developed this feature he never imagined that people would deploy it on DHCP servers and not change anything. He added that it was interesting to see how it worked because they assumed that DHCP servers would not allocate addresses and therefore would not run out of addresses in the pool. So it was cool to see that it just worked without doing that. He further said that for any connect, if the client was v4-only, and was just creating a v4-only UDP encapsulated thing that the local end of course had no idea what to do with, it had to be encapsulated in a v6 packet to send it. He added that he thought CLAT would be of no help but that it might work on Mac OS but not on other systems. He mentioned that IOS would never work until it went to v6 and that the Cisco people should be embarrassed about this. Veronika McKillop, UK IPv6 Council, asked which DHCP servers supported RFC8925. Ondřej replied that this was option 108 and that Kea, at least, did not support it. He added that there were two options. If you have enough addresses for the IPv4 pool, any server that allowed setting a custom option could be used. If there was any special support for this feature, the server would simply respond with option 108 and not allocate an address from the pool. He stated that he was not aware of such an implementation, as even Kea cannot be used in this way. He added that if the pool was empty, it would not even respond and that this was an interesting feature to implement for DHCP server producers. Jordi Palet Martínez, The IPv6 Company, said that it was worth mentioning that most of the CLAT implementations just used a default NAT64 prefix (the WKP). He added that there was only a single translation when using the CLAT and that two translations were only needed if there were no DNS64 (https://datatracker.ietf.org/doc/rfc8683). He said that even if your network was using NAT, NAT64 would also work well. Radek Zajic via Meetecho shared that the CLAT on iOS 16 also worked in apps requiring unix/raw sockets too (e.g. Hurricane Electric Network Tools can ping IPv4-only targets via the CLAT). Ondřej thanked Radek for his comment. Round Up and Thanks IPv6 Working Group Co-Chairs Raymond thanked all participants and the RIPE NCC and closed the session. For Internal Use Only

1 0

SLAAC and Broadcast RA
by Xuo Guoto 02 Nov '22

02 Nov '22

Dear WG, Apologies if this mail is off topic here. I have a question about BCOP-690 wrt SLAAC BCOP-690 says: Using a /64 prefix from a dedicated pool of IPv6 prefixes is the most common scenario and currently the best practice. A separate block of IPv6 space is allocated for the WAN links to the end customer CPEs, so that when CPE connects to the network and performs router discovery, a /64 prefix is used to number both ends of the connection. My understanding of this recommendation is that in an ISP scenario, each RG will be assigned a seperate /64 prefix for CPE WAN. The ISP BRAS or BNG is also expected to send periodic multicast RA. My question is how a multicast RA can be sent to a VLAN, in which each RG is expected to have a different /64 prefix? R-43 of TR-177 says: The BNG MUST be configurable to send multicast RA messages that do not contain a Prefix Information Option (PIO), in order to advertise the default router address. R-44 says: The BNG SHOULD support sending periodic and solicited multicast RA messages having the unicast MAC address of the RG or of the hosts behind a bridged RG (see RFC 6085 [26]). Reading these two together, is it correct to say that BNG should be sending out periodic multicast RA without PIO, but RA with PIO should be sent only as a reply to an RS message from RG, with multicast IPv6 destination address but with unicast link-layer address of RG. Thanks and Regards, X.

1 0