IETF 118 Prague
MBONED Agenda
Thurs, Nov 8, 2023
13:00-14:30 CET
Thursday session II, Amsterdam

Note-taker: Stig Venaas
Chat Log: https://zulip.ietf.org/#narrow/stream/mboned
Video log: https://www.youtube.com/watch?v=QJQire7vP6c

Notes taken in etherpad at https://notes.ietf.org/notes-ietf-118-mboned#

Text captured on 11/22/23:

Chairs Update:
Multicast telemetry doc passed wglc since 117
Looking for shepherd.
-Max volunteered to shepherd

YANG model draft. Sandy says received some comments from Tom and will update.
Sandy says redundant ingress failover doc should be ready for wglc
AMT yang model draft needs more review and comments.

Updates on MTTB drafts from Max. Dorms may be ready for wglc.

Multicast Extensions for QUIC:
Dino asking how it is working with keys.
Asking if retransmit as uc or mc if many clients don’t act. Basic idea is that draft doesn’t specify how to retransmit. Saying how unicast might be better if multicast failures.
Toerless asking if thought about NAT
Not sure if I got that right.
Someone (one of the authors?) saying that can kick client off multicast if too many failures.
A comment that many options how to do acks, may not need to ack every packet.
Mathias asking if this is just SSM. It seems that way.
Someone (think an author) saying this is harder for ASM with many sources.

Hitoshi asked about latency I think.
Greg pointing out that ASM or SSM is more about who does the source discovery, not whether many to many application.

Ice asking if routers do QUIC, I think.
Someone clarifying that lost packets are not retransmitted, but rather data is retransmitted.

Dino asking for more details on integrity.
Integrity frames are sent over unicast. Contains checksums of packets. Used to verify multicast packets.

Lucas Pardue Explaining about other approaches for integrity.

Francois Michel
Commenting on many to many, didn’t get it, but presenter says out of scope for now.

Juliusz
Asking about what crypto to use. Presenter says he can’t recommend a specific one.

Lenny
Cautions about adding complexity to address ASM. May make it harder to
deploy. Interdomain ASM was deprecated for a reason, recommend keeping
scope at SSM-only.

Dino
Asking if DH on multicast channels, rather than publishing shared key.
Juliusz mentions that 2 layers, encryption and authentication.

Toerless
Make it slow enough for the slowest guy?
Presenter says can be different channels with different rates.
Rely on clients joining slower channels if rate is too high for them.
Suggesting can keep it simple and just throw slowest guy out if needed.
Presenter says draft not specifying whether 1 or multiple channels. Not specified.
Toerless says a draft on multi channel behavior may be needed.

Francois Michel
Missed comment.

Juliusz
Asking if can do FEC for unreliable data. Says can work well.
Presenter says easy to do.

Gorry
Says congestion control in QUIC is very terse.

Hitoshi asking about use of different FEC for different clients I think. Seems not something they thought about.

Dino asking about cost of signatures. Whether to do it over just a few bytes or not. Maybe it can be cheap.

Lucas talking about how use cases are important.

Gorry Says a separate draft on congestion control could be useful to get help from CC experts. Says vital to get it right.
Greg asking if CC would be in QUIC itself or applications using it. Historically for multicast it’s been in the application.
Gorry not sure right now.
Toerless, there is a CC WG.

Scalability analysis comparison of multicast/BIER forwarding mechanisms:
(https://atlas.cs.uni-tuebingen.de/*menth/papers/Menth21-Sub-5.pdf)
Michael Menth, 20 min

Juliusz
Made some comment on worst case, I missed it.

Greg
Does not agree what presenter said about bier being bad for small groups. It may not generally be true.
It depends on whether receivers are in the same bitset.

Dino
Asking why multiple sets if few members. Greg is explaining.

No more questions.

TreeDN update by Lenny:
WGLC soon in mops wg
input from MBONED and PIM would be welcome