RIPE 70

Thursday, 14 May, 14:00-15:30

WG Co-Chairs: Rob Evans and Joao Damas

Scribe: Anand Buddhev

A. Administravia

Rob Evans opened the session. He apologised for the delay in distributing the minutes from the session at RIPE 69. Next he talked about chair rotation process. The WG chairs had sent the text of the proposal to the WG, and there were no comments. Joao Damas had said he was going to stand down. No candidates stepped forward for the position so Joao offered to stay on as co-chair. Rob asked if there were any objections and there were none. Joao was welcomed back as co-chair of the Routing WG.

Rob also apologised for the delay in publishing the WG charter. He said he would try to do this before the next meeting.

B. The Role of Analytics in Routing, Network Performance and SDN - Cengiz Alaettinoglu,Packet Design

The presentation is available at:
https://ripe70.ripe.net/presentations/152-RIPE-May-2015-Cengiz-Alaettinoglu.pdf

There were no questions or comments.

C. bgpdump2: A Tool for Full BGP Route Comparison - Yasuhiro Ohara, NTT Communications Corporation

The presentation is available at:
https://ripe70.ripe.net/presentations/144-ripe70-routing-wg-bgpdump2-2015-05-14.pdf

There were no questions or comments.

D. Latency IPv4 vs IPv6: Understanding the Difference - Alexander Azimov, Qrator Labs

The presentation is available at:
https://ripe70.ripe.net/presentations/170-Azimov.Latency-IPv4-vs-IPv6-Understanding-the-Difference.pdf

Cengiz Alaettinoglu, Packet Design, asked Alexander whether he had analysed a case where the IPv4 and IPv6 paths followed the same ASes, and whether there were differences in the delay.

Alexander said that he had done such comparisons and found a difference of about 50% in the delay.

E. RIPE NCC Updates on the Routing Information Service (RIS) - Colin Petrie, RIPE NCC

The presentation is available at:
https://ripe70.ripe.net/presentations/156-RIPE-NCC-updates-on-RIS-RIPE-70_final.pdf

Gert Doering, SpaceNet, said he liked the new system because it will show very short-lived announcements.

Joao Damas, Bond Internet Systems, asked whether the code for the state machine had been developed at the RIPE NCC, and whether the RIPE NCC would share it with the public.

Colin said that the code needed some more work and cleaning up, but that in principle, there was no objection to sharing it.

Martin Levy, CloudFlare, thanked Colin for the presentation. He said that the data from the RIS project is very useful, and there more there is of it, the more useful it is. He said that for the past two to three years, there have been no new peers, and he would like to see the RIS system grow. He stated his strong support for expanding the RIS system rapidly, and asked the RIPE NCC how many collectors it is willing to deploy, and whether they will be only in the RIPE region, or deployed globally.

Colin said that adding lots of peers has consequences for the back-end systems as well as the storage systems. He said that the RIPE NCC was taking a conservative approach in order to ensure that the back-end systems could cope with all the influx of new data. He also mentioned that RIPE NCC was discussing deploying new route collectors in co-operation with the other RIRs.

Romeo Zwart, RIPE NCC, added that at the moment, the RIPE NCC cannot make any firm commitments, but he thanked Martin for expressing a request for the RIS project to expand quickly.

Martin then asked for a show of hands for support of the expansion of the RIS project. Many hands were raised, expressing strong support for it.

John Bond, ICANN, said that he was very interested in the AS path feature that Colin talked about at the end of his presentation. He asked whether RIPE NCC will allow more peers to be connected to the multi-hop router collector, or deploy new multi-hop collectors for more capacity.

Colin replied that RIPE NCC was certainly keen to offer this feature to users.

Thomas King, DE-CIX, also thanked Colin for the presentation and expressed an interest in the AS path feature.

Z. AOB

Rob Evans did an update about the BoF session held on Monday evening. He said that the RIPE IRR has a lot of good quality data, but also data that is not authoritative. He talked about the different approaches discussed at the BoF session, and these were:

1. Cross-registry authorisation
2. Using RPKI signatures to generate route objects
3. Using RDAP
4. Drop the requirement that the AUT-NUM object holder authorise creation of a ROUTE object

He said a lot more discussion was required, and will be carried out on theDatabase WG mailing list. He invited interested parties to join the list and make their comments there.

Ruediger Volk commented that there had not yet been any discussion about how to distinguish good data from less reliable data. He suggested the idea of metadata in the database, to indicate whether the data had been authorised or not.