RIPE 71

Routing Working Group - RIPE 71
Thursday, 19 November 2015, 9:00-10:30
WG Co-Chairs: João Damas, Rob Evans
Scribe: Anand Buddhev

Administrivia

WG co-chair Joao Damas opened the session. Joao asked for comments about the
draft minutes of RIPE 70. There were no comments, and he declared them approved.

Isolario - Luca San

The presentation is available at: https://ripe71.ripe.net/presentations/12-pres.pdf

Joao Damas asked about the future of this project, and whether it will be supported. Luca answered that his institution plans to support it.

A Simple BGP - Alexander Azimov

The presentation is available at: https://ripe71.ripe.net/presentations/162-A-Simple-BGP.pdf

Warren Kumari (Google) commented that there is a draft by Jared Mauch and Job Snijders that achieves almost all the goals in this presentation. Alexander said that the draft came out around the same time he started his work.
He said the draft is about filtering route leaks, whereas his work is about avoiding them.

Randy Bush (IIJ) explained in detail about how this proposal is better than Brian Dixon's proposal, because it allows both parties to agree to a policy in the BGP OPEN message, rather than basing the route announcements on a marker applied to routes. This makes it more robust to accidental route leaks.

RIS Updates - Colin Petrie, RIPE NCC

Kaveh Ranjbar (RIPE NCC) commented to the WG that at the last meeting, the community asked for more RIS collectors to be deployed. He said that the RIPE NCC will certainly expand into areas that are currently not well-covered, such as Australia and New Zealand. He then asked the audience two questions and asked for people to raise hands:

1. What scale should the expansion be at? Would a few more collectors, say 15, be enough, or should be more like 50 to 100?
2. Whether RIPE NCC should deploy collectors in large ISP networks to get a different view.

Joao Damas asked whether this was to get a view inside an ISP network, or still get an external view but from an ISP's perspective. Kaveh said he meant the latter.

Kaveh asked for a show of hands, and there was little support for this option.

Turris - Ondrej Filip

The presentation is available at: https://ripe71.ripe.net/presentations/181-Turris-20151119-OF-RIPE.pdf

Mick O'Donovan (BT Ireland) commended Ondrej for this project. Joao Damas also added his commendation and hoped that other manufacturers would pick up this work, so that the quality of devices out on the Internet would improve.

Csaba Mate (NIIF/Hungarnet) asked whether there were any plans to release just the board, without the switch ports. Ondrej said that they have to produce large volumes to keep the costs reasonable, and so they don't plan on variants at this early stage. He added that the hardware design is open, and said anyone can produce a variant board by themselves.

Jen Linkova (Google) asked about IPv6 support. Ondrej said that it was no longer neceesary to talk about IPv6 separately. He implied that it was a given.

BGP collector Communities - Randy Bush and Emile Aben

The presentation is available at: https://ripe71.ripe.net/presentations/159-151119.ripe-rtg.pdf

Jen Linkova (Google) asked why Randy classified a hijack as an intention of misorigination and not as an intentional leak. Randy replied that it was possible to divide the hijacks into categories besed on intent or no intent.

Warren Kumari (Google) asked what Randy meant by "route processing time". Randy said he meant the time taken to start sending a full stream. He said they had eliminated the RIB to FIB timing, because that was the classic trap.

Peter Hessler (Hostserver) asked how Randy would deal with an organisation that has 32-bit ASN. Randy said he didn't know, and invited Peter to contribute to a draft.

Ruediger Volk (Deutsche Telekom) commented that he would be unhappy if someone tried to interfere with his organisation's internal communities setup. He said he would not mind if a route collector asked for specific community tags, as long as they described the whole topology, rather than just something specific to that collector. Randy said that trying to describe how to do the whole topology would be rather difficult. Ruediger then asked whether Randy was asking for these specific communities to sent just to router collectors, or to other peers to. Randy said that these communities needn't be sent to other peers, but one could, if one wished.

Ruediger then added that if he tagged his routes with communities and sent them to other peers, who then propagated them to Randy, he would see more interesting stuff. Randy replied that he didn't really want to see that. He said it was always possible to see which ASN a peer has because of the peering relationship.

AOB

Sander Steffann said that the issue of communities for 32-bit ASNs had come up in the Address Policy WG, and he asked them whether the routing WG should take this to the IETF. There was overwhelming support for this in the APWG, so the Routing WG will now take this to the IETF. He asked for questions. There were none, and there was loud applause.

Dario Rossi asked for help with the Anycasters project. He said if anyone could provide BGP feeds, it would be good, and invited people to talk to him.

Finally, Rob Evans announced that at the next RIPE Meeting, he would be stepping down as co-chair. He said that if anyone wished to take on this role, they should go and speak with him or Joao.