Markdown Version | Transcript | Session Recording

Session Date/Time: 11 May 2026 16:30

MOQ

Session: Media over QUIC (MoQ) Virtual Interim
Date: May 11, 2026

Summary

The session focused primarily on the proposal for "Filters" within the Media over QUIC transport, specifically regarding PR 1518 against draft-ietf-moq-transport. The group debated the technical complexity, utility, and architectural placement of range filters and "Top N" track selection filters. The meeting concluded with a brief temperature check and survey regarding the ongoing "Joining FETCH" design controversy.

Key Discussion Points

Filters and PR 1518

Mo Zanaty presented Filters and MOQT Issues and PRs, outlining a reduced scope from previous proposals.

• Filter Types:
  • Range Filters: Targeting Object ID, Subgroup, Priority, and Property types. These allow for use cases like video keyframe scrubbing, layer extraction, and priority-based logging.
  • Track Selection (Top N): Allows subscribers to request the "Top N" active tracks (e.g., loudest speakers in a conference) based on a property.
• Technical Design: The PR introduces setup options to limit computational overhead (Max filter ranges and Max tracks selected). It also defines a state machine for track selection (Unknown, Selected, Deselected) and utilizes delta encoding.
• Implementation Reports:
  • Mo Zanaty noted successful implementations in Quicker Laps (used in an NAB demo), Open MoQ (MockX), and Cloudflare (MockRS).
  • Tim Evens reported that while implementation required refactoring, performance hits were minimal (less than 1% in flame graphs).
  • Alan Frindell raised concerns regarding the complexity of "self-exclusion" (ensuring a publisher doesn't receive their own track in a Top N selection) and the state management burden on relays.
• Architectural Placement: A significant portion of the debate centered on whether these features belong in the core draft-ietf-moq-transport or an extension draft.
  • Pro-Core: Cullen Jennings, Suhas Nandakumar, Mo Zanaty, and Will Law argued that filters are fundamental to MoQ’s value proposition (network-side data reduction) and that splitting them could lead to synchronization issues with SUBSCRIBE_NAMESPACE.
  • Pro-Extension: Alan Frindell and Ian Swett suggested that the complexity and remaining "unbaked" aspects (like DDoS mitigation) warrant placing them in a separate document to avoid delaying the core transport draft.
• DDoS and Security: Ian Swett highlighted potential amplification attacks where a relay could be forced to perform unbounded filtering work. Mo Zanaty and Tim Evens noted that while setup limits help, more robust text is needed in the security considerations.
• Group Boundaries: Victor Vasiliev and Will Law noted that for media use cases, filter changes (like switching SVC layers) should ideally occur at group boundaries to avoid truncated data.

Joining FETCH

Alan Frindell presented Joining FETCH Temperature Check.

• The chairs noted there are currently approximately six competing proposals to address "Joining FETCH" (the mechanism for joining a live stream and fetching past data).
• Alan Frindell issued a three-question survey to the group to gauge whether the current design meets use cases, if participants can "live with" the current control-plane/data-plane split, and whether any new mechanism should augment or replace the existing one.

Decisions and Action Items

Poll Results

Magnus Westerlund conducted two informal polls to gauge the room's sentiment on including filters in draft-ietf-moq-transport:

1. Merge Range Filters into core draft?

   • Yes: 7
   • No: 3
   • No Opinion: 1
   • Result: Indicates a rough consensus in favor.

2. Merge Track Selection (Top N) Filters into core draft?

   • Yes: 7
   • No: 7
   • No Opinion: 2
   • Result: No consensus; significant division remains.

Action Items

• Mo Zanaty: Rescrub the 17 open issues related to the Filters PR and coordinate with openers to close resolved items.
• All Participants: Respond to the "Joining FETCH" survey on the mailing list by the following Wednesday (May 20th).
• Chairs: Summarize the poll results and the path forward for Filters on the mailing list.

Next Steps

• The next Virtual Interim is scheduled for Tuesday, May 26, 2026.
• Discussion will continue on the mailing list regarding whether to split the Filter PR into two (Range filters vs. Track selection) based on the poll results.
• Preparation for the London meeting (June) including interop details.

Related Documents

draft-ietf-moq-transport