Markdown Version | Transcript | Recording 1 | Recording 2 | Session Materials

Session Date/Time: 17 Mar 2026 01:00

MOQ - IETF 125 Meeting Minutes

Summary

The MOQ working group met for its first session at IETF 125. The session focused on significant updates to the core transport protocol, refinement of the URI scheme, and various security-related drafts, including end-to-end encryption and authentication mechanisms. The group also discussed dynamic track switching for relay-side adaptive bitrate (ABR) control.

Key Discussion Points

MOQT Update

Presenter: Alan Frindell Slides: MOQT Update

• Wire Image Overhaul: draft-ietf-moq-transport underwent significant changes between versions -15 and -17. Notable changes include replacing the single bidirectional control stream with a pair of unidirectional streams for setup, moving requests to their own bidirectional streams, and introducing MoQ-specific varints.
• Renaming and Properties: SETUP_PARAMETERS were renamed to SETUP_OPTIONS. The protocol now distinguishes between "Track Properties" (end-to-end) and "Message Parameters" (hop-by-hop).
• Control/Data Plane: Improvements were made to SUBSCRIBE_NAMESPACE (now supports discovery), delta encoding for efficiency, and the ability to mix streams and datagrams within a single track.
• Issue Tracking: To manage complexity, major features like Filters and Dynamic Track Switching are being tracked in separate repositories/drafts before merging.
  • Discussion: Colin Perkins raised concerns regarding IPR and archiving if development happens in individual repositories rather than the IETF-managed GitHub. Martin Duke (Chair) suggested that significant features should be adopted as separate WG drafts to ensure proper IPR coverage.

moqt:// URL Scheme

Presenter: Alan Frindell Slides: moqt://

• Transport Selection: The group discussed the moqt:// URI scheme. Previously, +q (QUIC) and +wt (WebTransport) suffixes were proposed.
• Decision: There was a strong sense of the room, supported by Ted Hardie and Martin Duke, to remove the +q and +wt suffixes. Transport selection should be handled via ALPN and protocol negotiation rather than the URI scheme.
• Fragments: A discussion ensued regarding the use of fragments (#) for application-level instructions (e.g., in draft-ietf-moq-msf). Ted Hardie and Harald Alvestrand expressed concerns about deviating from standard URI fragment semantics (usually tied to MIME types), while Will Law and Mike Bishop argued for the pragmatic use of fragments to pass information to the client application without involving the server.

End-to-End Secure Objects

Presenter: Colin Perkins Slides: Secure Objects Draft: draft-ietf-moq-secure-objects

• AAD and Key Derivation: Discussion on what data should be included in the Authenticated Additional Data (AAD). Victor Vasiliev and Mo Zanety suggested binding keys to the track name rather than the namespace to prevent nonce reuse across tracks.
• Track Properties Security: The group debated how to protect track properties (integrity and/or encryption).
  • Option 1: Move sensitive properties to the first object.
  • Option 2: Authenticate properties by including them in every object's AAD (suggested by Richard Barnes to hash them into the keys instead).
  • Option 4: Define a specific encryption scheme for control messages.
• Outcome: Colin Perkins requested clear requirements from users of track properties to determine which protection level is necessary.

Privacy Pass Authentication

Presenter: Thibault Cholez Slides: Privacy-Pass-MOQT Draft: draft-ietf-moq-privacy-pass-auth

• The draft introduces unlinkable tokens for anonymous subscriptions.
• Reverse Flow: Allows a relay to reissue tokens after initial bootstrapping to maintain long-lived, privacy-preserving sessions.
• Discussion: Altanay asked about the impact on relay chains. Thibault Cholez noted that each relay would likely need its own token to maintain privacy and prevent linking across hops.

Application-Agnostic DPoP Proof for C4M

Presenter: Suhas Nandakumar Slides: Application-Agnostic DPoP Proof for C4M Draft: draft-ietf-moq-c4m

• The draft addresses the need for sender-constrained tokens (Demonstrating Proof-of-Possession or DPoP) in MoQ. Current DPoP is HTTP-specific.
• Richard Barnes noted that DPoP is critical for MoQ to prevent relays from replaying bearer tokens to impersonate clients.
• Coordination: Suhas Nandakumar and Martin Duke noted that work needs to be coordinated between the MOQ and OAuth working groups.

Dynamic Track Switching (DTS)

Presenter: Will Law Slides: Dynamic Track Switching for relays - v2

• DTS enables server-side/relay-side ABR by allowing a client to subscribe to a "set" of tracks and letting the relay switch based on throughput.
• Refinements: Will Law proposed removing the "waiting/timeout" timer and the activation parameter to simplify the implementation.
• Throughput Fractions: Proposed a way for a client to tell a relay to use only a portion of the connection's bandwidth for a specific switching set (useful for multi-view scenarios).
• Discussion: Kutish Remable and Suhas Nandakumar discussed the challenges of publisher-controlled switching in conferencing scenarios where the client's layout determines the needed bitrate.

Decisions and Action Items

• Decision: Remove transport-specific suffixes (+q, +wt) from the moqt:// URI scheme in draft-ietf-moq-transport.
• Action Item: MoQ Editors and Chairs to resolve IPR management for features being developed in external repositories (e.g., Filters, Switch).
• Action Item: Chairs and Suhas Nandakumar to coordinate with the OAuth WG regarding the progression of draft-ietf-moq-c4m.
• Action Item: Colin Perkins to expand the Secure Objects issue tracker to include the three proposed varint encoding solutions.

Next Steps

• Draft-18 Target: Approximately 8 weeks from IETF 125, serving as the next Interop target.
• Interim Meetings: A series of virtual interims are scheduled for Mondays (beginning April), and a hybrid interim will be held at Cloudflare’s London offices in June.
• Dynamic Track Switching: The group will continue to iterate on the design as a separate draft before deciding whether to merge it into the core transport specification.

Session Date/Time: 19 Mar 2026 01:00

MOQ

Summary

The MOQ working group met at IETF 125 in Shenzhen to discuss the progression of the core transport protocol, streaming formats, and production implementation experiences. Key highlights included a report on production usage at Alibaba, updates on the MOQT Streaming Format (MSF), and a deep dive into filtering mechanisms and transport-level issue resolution for draft-ietf-moq-transport.

Key Discussion Points

Hackathon and Interop Status

• Mike English (Cloudflare) reported on the interop test runner, which now includes automated tests for various client and relay implementations.
• Discussion on Versioning: Alan Frindell suggested that while Draft 17 is available, the working group should target Draft 18 for the next major interop milestone (London interim). This is due to the significant changes between Draft 16 and 17. Mike English will initiate a list discussion to confirm this transition plan.

MOQT Streaming Format (MSF) and CMSF

• Will Law (Akamai) presented MSF & CMSF update - Shenzhen - v1 regarding draft-ietf-moq-msf and draft-ietf-moq-cmsf.
• Renaming: "Warp" is now formally MSF; "CARP" (CMAF-compliant) is now CMSF.
• URL Format: A fragment-based scheme (e.g., #msf:) is proposed to identify namespace and track metadata.
• New Features: Added support for variable substitution in catalogs (for caching/personalization), media timeline tracks for scrubbing, and SCTE-35/caption tracks.
• Discussion: Benjamin raised questions regarding the necessity of authorization for receiving tracks if content is encrypted. Will Law clarified that authorization (using CAT or Privacy Pass) controls access to the relay resources, not just the content.

MOQ Production at Alibaba

• Minghui presented MOQ production at Alibaba, highlighting deployments in voice search, cloud rendering, and AI assistants.
• Performance data showed significantly lower latency compared to WebRTC and WebSocket.
• New Draft Proposal: Proposed a multimodal feedback mechanism (using a dedicated MoQ track) to allow clients to report delivery status and buffer levels to the publisher for ABR and AI inference adaptation.

MOQT over QMUX

• Suhas Nandakumar presented MOQT-over-QMUX.
• The goal is to provide a TCP fallback for MOQ in environments where UDP/QUIC is blocked.
• Negotiation: Discussion focused on whether to use a specific ALPN or a layered approach similar to WebTransport. Alan Frindell expressed a preference for a single MoQ ALPN that implies the QMUX version.

Low Overhead Container (LOC)

• Mo Zanaty provided an update on LOC (draft-ietf-moq-loc).
• Metadata: Metadata has been split into public properties (in the MoQ header) and private properties (within the MoQ payload) to better support encryption and relay opacity.
• Timescale: Added a timescale property to define timestamp denominators (e.g., 90kHz for video).

Filters

• Mo Zanaty presented Filters, focusing on PR 1518.
• Track Selection Filter: Introduced a "Top N" filter for namespaces, enabling use cases like active speaker switching at the relay.
• Relay Protection: Discussed the inherent asymmetry of filters (simple client request vs. complex relay processing). New error codes like conflicting-filters and namespace-too-large were introduced to allow relays to protect resources.
• Discussion: Alan Frindell and Ian Swett noted concerns regarding recursion and state management. Mo Zanaty clarified that relays could explicitly signal deselection using forward=0.

Transport Issues (draft-ietf-moq-transport)

• Alan Frindell led a discussion on MOQT PRs and Issues.
• Variant Encodings: A debate occurred regarding whether to enforce minimal encoding for the new MoQ variable-length integers (VI64).
  • Poll: "Should minimal encoding be a MUST?"
    • Result: 4 in favor, 22 against.
  • Poll: "Should 7-byte variants remain disallowed (for extensibility)?"
    • Result: 15 in favor, 5 against. This will be followed up on the mailing list.
• VI64 vs V64: The group agreed to use the name "VI64" for the variant type.

Decisions and Action Items

• Transport: The REQUEST_OK message will be updated to include a properties field to maintain consistency with other track-related messages.
• Transport: Variant encoding requirements will remain a "SHOULD" for minimal encoding, following the behavior of QUIC.
• Interops: Mike English will coordinate on the mailing list to target Draft 18 for the London Hackathon.
• Naming: The variable-length integer type will be officially named VI64.

Next Steps

• Interim Meeting: Scheduled for March 30, 2025, at 16:30 UTC (to include a demo of server-side ABR).
• London Interim: A hybrid interim meeting is planned for June in London.
• Draft Progression: Editors will incorporate the decided changes into draft-ietf-moq-transport and draft-ietf-moq-loc.

Related Documents

draft-18, draft-ietf-moq-c4m, draft-ietf-moq-cmsf, draft-ietf-moq-loc, draft-ietf-moq-msf, draft-ietf-moq-privacy-pass-auth, draft-ietf-moq-secure-objects, draft-ietf-moq-transport