RIPE 69

Open Source Working Group - RIPE 69 Draft Minutes
Wednesday, 5 November, 9:00-10:30
WG Chairs: Martin Winter, Ondrej Filip
Scribe: Anand Buddhdev

A. Admin matters

The working group chairs, Ondrej Filip and Martin Winters, welcomed the participants. The agenda was reviewed and some presentations were shuffled to allow a presenter to leave early.

There were no comments on the minutes of the WG session of RIPE 68.

There were no outstanding action items.

B. Lightning Update

Sara Dickinson gave a presentation about Hedgehog.

The presentation is available at:
https://ripe69.ripe.net/presentations/76-Hedgehog_RIPE69.pdf

Ondrej Sury asked if they could provide a Flashless version. Sara said that in the upcoming 2.0.0 version, there will be a choice between Flash and SVG.

C. Using OpenSource to develop OpenSource

- Ondřej Surý, CZ NIC

The presentation is available at:
https://ripe69.ripe.net/presentations/77-OS-20141105-RIPE69.pdf

Peter van Dijk (PowerDNS) commented that not all projects have resources for running their own gitlab, and recommended the use of Github, Travis and Coverity.

D. Use cases of ExaBGP, How ExaBGP Got Where It Is and What Is Next

- Thomas Mangin

Kaveh Ranjbar (RIPE NCC) thanked Thomas for this project, and said that the RIPE NCC is going to start deploying new RIS (Routing Information Service) servers based on ExaBGP, and will provide updates soon. He then asked Thomas if there would be any support for RPKI-RTR (RFC 6810) in ExaBGP. Thomas said he will not do it himself, but he will accept contributions.

E. Lightning updates

Lua Policy Engine

- Peter van Dijk

The presentation is available at:
https://ripe69.ripe.net/presentations/55-powerdns-ripe69.pdf

There were no questions.

Bird Update

- Ondrej Filip

The presentation is available at:
https://ripe69.ripe.net/presentations/79-BIRD-RIPE-Update.pdf

Martin Winter asked how RFC compliance testing was done in the code. Ondrej said there were not many formal methods. He said they have a test network that changes often, and is built badly by design. They also involve beta testers. He said it was not easy to test it in some known ways, so most testing is done with live networks.

Quagga Update

- Paul Jakma

The presentation is available at:
https://ripe69.ripe.net/presentations/85-ripe69-os-quagga-update.pdf

Nick Hilliard (INEX) commented that Cumulus Networks had posted many patches lately, including a fix for a BGP renegotiation problem (causing some peers to just fail to connect after a restart of Quagga). He asked whether Paul could prioritise merging these patches in, especially as they depend on several other outstanding patches. Paul said he wasn't sure which patches they were, and asked Nick to indicate which patches he was talking about.

Nat Morris (Cumulus Networks) said that their patches are on GitHub, and one can compile the code, or use their DEB packages directly. He said Cumulus Networks would be happy to help integrating these patches.

NSD 4.1 Update

- Jaap Akkerhuis

No presentation was uploaded.

There were no questions.

F. Open source working group chair terms and conditions

Martin Winter and Ondrej Filip introduced the issue of chair selection, and invited the group to give their input.

Niall O'Reilly referred the group to Nigel Titley's ideas in the Database Working Group session. Martin asked Niall to give more details. Niall explained that when a chair needs to step down, there's a nomination to the group, followed by a negotiation phase to find a candidate. If this doesn't work, names are drawn out of a hat. Martin thought it may not be the best choice and said that the negotiation phase was a strange idea.

Shane Kerr said that a simple vote by show of hands is easy, and he prefers it over other ideas.

Gilles Maassen said that RIPE working group work is done by consensus, and so he prefers chair selection by consensus. This would have the advantage of not excluding members who were not in the room.

Leslie Carr suggested the use of IRC chat and IRC "hands" for voting. Martin said that most RIPE WG decisions are taken by voting at a meeting; people not present don't have a right to vote.

Benedikt Stockebrand (speaking as incoming chair of the IPv6 WG) said that consensus is the best. If possible, it should always be used to select a chair. However, if consensus fails, there should be a policy in place (vote, random draw) as a last resort to stop things getting out of control.

Peter Koch (DENIC) said that a vote was probably not the right thing to do. He said that for a proper vote, an electorate is required, and that without a proper setup, an IRC vote could easily be "gamed". He agreed with Gilles Maassen and said that building consensus was important, and that if a tie needed to be broken, then drawing a name out of a hat, rather than a vote, was the fairest method.

Kostas Zorbadelos (via chat) asked about introducing candidate profiles. Ondrej Filip said candidates would be expected to introduce themselves on the mailing list about a month before the selection. Martin reiterated the same.

Martin then asked or opinion about the number of chairs, and whether it should be two or three. Benedikt Stockebrand said that having two could lead to a deadlock about opinion. Having three makes it easier, because the one who opinion differs from the other two's can perhaps more easily change his mind. Ondrej said that currently having two chairs has led to disagreements, and Martin jokingly added that he usually wins.

Martin asked people to think about volunteering for the position of chair, and submit their intention a month before the next RIPE meeting.

Ondrej Filip thanked everyone for attending.

Meeting ended.