CHAIRS: Jonathan Lennox, Marius Kleidl
IETF 123 Agenda
Location: Madrid, Spain
Session: IV
Room: Patio 1
Date: Tuesday, 22 July 2025
Time: 17:00 - 19:00 Madrid time
Preliminaries (Chairs, 10 min)
Note Well, Note Takers, Agenda Bashing, Draft status
SDP Offer/Answer for RTP over QUIC (RoQ) (Spencer Dawkins, 15 min)
https://datatracker.ietf.org/doc/draft-dawkins-avtcore-sdp-roq/
Bundle PR
Harald: some RTP sessions refer to multiple SSRCs
Nils: Bundle is more about multiple RTP flows on the transport not just
RTCP muxed with RTP.
Jonatahn: if BUNDLE isn't used each RTCP feedback would need to be
transported on a separate channel
Magnus: I think you need to fix this in SDP. The middlebox can not be
expected to magically untangle this.
ICE
Nils: Alternative is to have separate SDP m lines for QUIC vs
traditional ICE candidates.
Magnus: ICE/STUN process doesn't need to care about encapsualtion. But
you need to signal the different encapaulation.
Harald: QUIC can't connect without a servername. ICE doesn't care about
server names. We are redefining P2P QUIC here. SVAP says you need to use
protection, but in case of QUIC handles the protection to you drop it.
Jonathan: should a single m-section be able to contain a fallback from
RoQ?
Nils: the question is if you know if the answers can handle RoQ or not.
If a fallback is needed separate m-sections with different transports
(one RoQ, one with non-RoQ) would be the safer option.
Mo: Is anything missing for P2P ROQ? Answer is no.
Spencer: please consider call for adoption
Jonathan: the draft has already been treated almost as a WG draft
Magnus: what are the proporties of this avatar? Is it required to
arrive?
Mo: do all clients have already a full mesh sets and you only exchange
in expressions.
Hyunsik: Yes only changes are transfered.
Mo: there is already a vertex manipulation proposal from long time ago.
See draft-jennings-dispatch-game-state-over-rtp.
Jonathan: is the payload format also going to be standardized in IETF?
Hyunsik: not sure yet.
Mo: Why is timestamp insufficient? Erik: Need spatial layer ID as well.
Also, prefer consecutive without gaps.
Jonathan: frame number ID is not a good name choice.
Would you receive feedback from multiple clients or only a single
point?
Erik: yes the SFU needs to agregate the feedback
Wait until November for a call for adoption.
Ready to be put on the list for adoption. One support, no objection in
the room.
The first WGLC didn't show a lot of interest.
Another WGLC?
Gorry: please speak up if you are interested
Erik: volunteers to review
Youngkwon: also offered to review