Minutes IETF111: babel
minutes-111-babel-01

Babel WG
VIRTUAL - Tuesday, 26 July 2021
21:30-22:30 (UTC) Room 4
14:30-15:30 (OS Pacific Time)

Chairs: Russ White (Juniper)
Donald Eastlake (Futurewei)

Area Director: Martin Vigoureux (Nokia)

Minutes: Taken by Barbara Stark, edited by Donald Eastlake

Agenda (amount of time originally scheduled, not time taken)
 3 min. Administrativia (scribes), Agenda Bashing, Chairs
 7 min. Status, Milestones, Chairs
           draft-ietf-babel-yang-model-10
15 min. Let’s revive Babel-RTT, Juliusz Chroboczek
           draft-ietf-babel-rtt-extension-00
25 min. Wi-Fi Addressing, Transit, Multicast, Mesh, Donald Eastlake
 3 min. Wrap-Up, Chairs

Barbara Stark volunteered to take minutes.

A youtube video of the seesion is at
https://www.youtube.com/watch?v=uM1McPr4oK8

Administrivia, Status, Milestones
=================================
Chair Slides: 

Donald Eastlake (Futurewei, Chair): Donald went through the chair
slides. On slide covered draft-ietf-yang-model.

Mahesh Jethanandani (Kloud Services): Mahesh spoke to the slide on the
YANG model. It is almost done with IESG review. One last set of
comments from Ben will be addressed after this IETF meeting and a
revised draft uploaded.

Donald: Despite by best efforts to add some not-yet-accomplished
future milestones, the working group was only agreeable to adding one
and we have done that so it is still true that we have accomplished
all our milestones.

Let’s Revive Babel-RT
=====================
Slides

Juliusz Chroboczek (IRIF, University of Paris) presented.

After the slides he asked for comments/suggestions:

Donald: I think you should use different type values for the sub-TLVs,
as you suggest. And that experimental is a good category.

Juliusz: There is some argument for keeping it compatible with the
current implemented base. There is less liberty to the implementor to
do the wrong thing. There are good reasons to use the same TLVs and
good reasons not to.

Donald: Compatibility with running code is a strong argument.

David Schinazi (Google, jabber): Since both current or new sub-TLVs
work, I would suggest keeping what is in production today

Donald: There has been considerable interest in this in the WG, both
in meetings and on the mailing list, so it seems like a good idea to
continue work on it.

Wi-Fi Addressing, Transit, Multicast, Mesh
==========================================
Slides

Donald Eastlake presented.

Juliusz: Going back to slide 12. Why are there up to 4 addresses
instead of just 2?

Donald: One set is link local, the other in "end-to-end". You normally
have 4 addresses, it is just that usually you have two Layer 3
addresses and two Layer 2 addresses. So the addresses just aren’t of
the same kind. Since 802.11 is from the IEEE 802 world, it only uses
MAC addresses. The "Distribution System" show on slide 6 does routing
but is deliberately underspecified in the 802.11 standard. You can
implement the Distribution System routing however you want, with Layer
2 or Layer 3 mechanisms, because standard doesn’t specify.

Juliusz: Radia Perlman has a beautiful talk that says "~routing should
be done at Layer 3 but people want to do it at Layer 2 so I designed
TRILL as a way that they can at least do Layer 2 routing properly~".
Why not do routing at Layer 3 where it belongs?

Donald: TRILL is really more like Layer 3 routing (IS-IS) but using
flat MAC addresses rather than hierarchically structured IP addresses.

Juliusz: If a DS routes at Layer 2, it will cause a lot of extra
complexity. Why do people want to rqoute at Layer 2? Layer 2 is
unnecessarily complicated. I think there are two reasons. Layer 3
configuration is harder. The other reason is it also allows for easier
mobility to do routing at L2.

Donald: The 5G core is also an example of L2 routing.

Juliusz: My hypothesis is that you have 4 addresses because you’re
doing routing. It should be possible to do the DS at Layer 3,
similarly to what the LTE core network does. I would like to point
people to git clone https://www.irif.fr/~jch/software/sroamd.git . I
hope to get back to this.

Closing
=======
Donald: We’re out of time. See you at the next meeting and on the
mailing list. If there is enough interest in a topic, we can schedule
an interim meeting to make more rapid progress.