AVTCORE

Summary

The AVTCORE working group met at IETF 125 to discuss various RTP payload formats and protocol enhancements. The session covered the recently adopted APV and Avatar (ARF) payload formats, the status of H.265 for WebRTC, and SFrame packetization. Significant discussion was held regarding a new generic RTP Frame Acknowledgement mechanism. Additionally, two new proposals aimed at accelerating WebRTC connection setup (SPED and SNAP) were presented to determine the appropriate venue for future work.

Key Discussion Points

RTP payload format for APV

Presenter: Suhyeok Jeong Slides: RTP payload format for APV Draft: draft-ietf-avtcore-rtp-apv

Suhyeok Jeong provided an overview of APV (Advanced Professional Video), recently published as RFC 9924.
The draft specifies two packetization modes: Simple Mode (basic fragmentation) and Low Delay Mode (aligning tile boundaries with payload starts for immediate decoding).
Mo Zanaty recommended renaming fields like "payload type" within the APV header to avoid confusion with the standard RTP header's Payload Type (PT) field.
Jonathan Lennox suggested investigating the need for an extended sequence number, noting that at the high bitrates expected for professional video, the 16-bit RTP sequence number might wrap around too quickly (referencing RFC 4175).

H.265 Profile for WebRTC

Presenter: Philipp Hancke Slides: H.265 Profile for WebRTC Draft: draft-ietf-avtcore-hevc-webrtc

Philipp Hancke announced Jianlin (Intel) as a new co-author. The implementation has shipped in Chrome 136 (hardware) and Safari 18.
Mo Zanaty clarified that WebRTC implementations of HEVC use in-band parameter sets on keyframes.
The draft is considered feature-complete.

RTP Payload Format for Avatar Representation Format (ARF) Animation Stream

Presenter: Suhyeok Jeong Slides: RTP Payload Format for Avatar Representation Format (ARF) Animation Stream Draft: draft-ietf-avtcore-rtp-avatar

The draft was updated to align terminology with MPEG specifications.
The authors are waiting for the MPEG Avatar specification to reach the Final Draft International Standard (FDIS) stage before finalizing the RTP payload header sections.

RTP Frame Acknowledgement version 2

Presenter: Gurtej Kanwar Slides: RTP Frame Acknowledgement version 2 Draft: draft-kanwar-avtcore-rtp-frame-acknowledgment

Gurtej Kanwar presented a mechanism for receivers to acknowledge decoded frames, allowing senders to reference known good frames for recovery instead of sending large intra-frames.
Mo Zanaty and Stefan (Lohning) questioned if a truly codec-agnostic mechanism is possible, noting that most recovery requires specific codec-level bindings (similar to RPSI). The authors argued the wire format can be generic while the logic remains in the implementation.
Altanai Sisodia raised concerns regarding potential amplification attacks if a receiver requests feedback for a massive range of frames. The authors noted that RTCP is authenticated, but agreed to add guidance on memory limits for the status vector.
Jonathan Lennox raised questions regarding how the mechanism handles Scalable Video Coding (SVC) when multiple layers share a timestamp. Eric (Rescorla) clarified that frame IDs would uniquely identify decodable units regardless of timestamps.

SFrame RTP Packetization

Presenter: Yoann Slides: SFrame RTP Packetization Draft: draft-ietf-avtcore-rtp-sframe

Introduced a "T" bit to distinguish between raw (WebRTC encoded transform) and pre-packetized content.
Added per-SSRC key derivation and integration with SFrame ratcheting.
Implementations are progressing in libdatachannel (Matter/Chips) and libwebrtc.

STUN Protocol for Embedding DTLS (SPED)

Presenter: Justin Uberti Slides: STUN Protocol for Embedding DTLS

This proposal embeds DTLS handshake packets within STUN attributes during the ICE connectivity check phase, potentially saving one or more RTTs.
Harald Alvestrand noted that Google has implemented a similar mechanism in Chrome and is interested in standardization.
Gorry Fairhurst (TSVWG Chair) suggested coordination between AVTCORE and TSVWG to determine where this work should reside, given it touches transport and application layers.

SCTP Negotiation Acceleration Protocol (SNAP)

Presenter: Philipp Hancke Slides: SCTP Negotiation Acceleration Protocol

Proposes carrying the SCTP INIT chunk in the SDP offer/answer to eliminate the SCTP handshake delay for WebRTC Data Channels.
Jonathan Lennox and Harald Alvestrand discussed potential issues with parallel forking and provisional answers (PRANSWER).
Magnus Westerlund suggested that while SCTP experts (TSVWG) should review it, the core complexity is in the SDP mapping, which fits AVTCORE or the former MMusic scope.

Decisions and Action Items

[draft-ietf-avtcore-hevc-webrtc]: Philipp Hancke to submit a new version addressing minor nits. Once submitted, the chairs will initiate a Working Group Last Call (WGLC).
[draft-ietf-avtcore-rtcp-green-metadata]: Currently in WGLC; members are encouraged to review and provide feedback.
[draft-kanwar-avtcore-rtp-frame-acknowledgment]: The chairs will issue a Call for Adoption.
[draft-ietf-avtcore-rtp-apv]: Suhyeok Jeong to address feedback regarding field naming and sequence number wraparound.

Next Steps

SPED and SNAP: Chairs (Jonathan Lennox and Marius Kleidl) will coordinate with TSVWG chairs and the AD to determine the appropriate working group for these proposals.
ARF: Update expected once MPEG FDIS is available.
SFrame: Targeting WGLC by the end of the year, following further implementation feedback.
Interim Meeting: The chairs will plan an interim session, likely around May, to maintain momentum on active drafts.

Automatic IETF Minutes