Session Date/Time: 09 Dec 2025 20:00

CELLAR

Summary

The CELLAR Working Group met to discuss the status of existing and new documents, including the progression of draft-ietf-cellar-tags and draft-ietf-cellar-codec, and timelines for previously paused drafts. Significant time was spent reviewing outstanding RFC errata reports for Matroska (RFC 8794) and EBML (RFC 8797), determining their appropriate classification. The group also discussed the formal process for adopting draft-ietf-cellar-matroska-v5 as a working group document and methods for public engagement with its development. An inquiry regarding the registration of new audio codecs was also addressed.

Key Discussion Points

Administrative Items
- The IETF NoteWell was reviewed.
- The October interim meeting minutes were approved with no corrections.
- The meeting schedule for 2026 was discussed and approved, canceling two meetings that would overlap with IETF sessions and moving the November meeting to the second Tuesday after American Thanksgiving. AD approval for interim meeting requests is no longer required; requests are entered directly into the tracker.
Working Group Documents Status
- draft-ietf-cellar-tags: Revision -20 was submitted and is currently in ISG evaluation. It is scheduled for the ISG telechat ballot in early January 2024.
- draft-ietf-cellar-codec: An update was submitted to correct a reference and is now in requested publication status.
- Expired Drafts (chapter-codex, ffv1-v4):
  - chapter-codex: The plan is to return to this draft in 2026. The existing support in libdvdread for DVDs should be sufficient.
  - ffv1-v4: Development remains paused but is expected to resume in 2026. Key areas of work include implementing Michelle Niedermeyer's suggestions for storing raw bits in high bit-depth scenarios and incorporating floating-point support, which is already present in FFmpeg code. Benchmarking and specification writing are needed. Funding for FFv1 work via FFmpeg was mentioned as a possibility for writing the specification.
  - GPU acceleration for FFV1 v3 is being optimized, with insights expected to inform FFV1 v4.
- Milestone Dates: A discussion was held regarding removing specific dates from working group milestones in the data tracker if there is no external dependency requiring published dates. It was noted that ffv1-v4 and chapter-codex drafts currently do not have external pressure for specific publication dates, making dateless milestones a viable option.
RFC Errata Reports (RFC 8794 - Matroska, RFC 8797 - EBML)
- Matroska Errata (RFC 8794):
  - Errata ID 8615 (Matroska: AttachmentLink MaxAccurs): This errata proposes removing a MaxOccurs constraint, clarifying that an attachment can have multiple links. It was characterized as an editorial mistake/copy-paste error. The sense of those present was to recommend it be classified as "verified."
  - Errata ID 8616 (Matroska: Codec Delay): This errata clarifies the specification for codec delay and other related elements to tighten interpretations and ensure consistent timestamp computation. While a clarification, it was noted that the AD might consider it a protocol change, potentially leading to a "held for document update" classification. If classified as "held for document update," it could be addressed when Matroska v5 (which would obsolete v4) is published.
- EBML Errata (RFC 8797):
  - A general review of eight outstanding errata for RFC 8797 (EBML) was encouraged.
  - Errata regarding element ID 0x80: This element ID was marked as reserved in the EBML specification but is already in use in Matroska. The proposal is to remove the reserved status. Given its existing use in Matroska, the sense of those present was to recommend this errata be classified as "verified." The IANA EBML ID registry currently lists 0x80 as reserved.
Matroska V5 (draft-ietf-cellar-matroska-v5)
- There was a consensus to proceed with submitting a first draft. The draft is already built, refined, and includes additions like Zstandard, which has been tested in VLC.
- The process for adopting a new working group document was discussed. The CELLAR charter text already supports Matroska v5, so no re-chartering is needed. However, the formal adoption process with the AD still needs to be followed, starting with Spencer contacting the AD.
- Methods for engaging the community and exposing the work-in-progress Matroska v5 on GitHub were discussed, including potential nightly/weekly builds of the XML/HTML outputs. The current process involves checking artifacts from pull requests, but these might be transient.
Other Business
- An email inquiry about registering audio-vivid 3D audio codecs for use in Matroska was forwarded to the mailing list. The challenge is that the codec-id registry is not yet an RFC, which affects the standard registration process. Steve agreed to respond to the inquiry.
- Signaling Timecode Source: A pull request (PR) exists for signaling the source of time codes in BlockAdditional elements, including vocabulary for source types. This PR will be reviewed offline.
- Late Timecode Signaling: A question was raised about writing BlockAdditional elements when timecode signaling is known late (e.g., only after a video frame), and insufficient space was reserved in the Tracks element at the beginning of the file. Options discussed included reserving ample space, voiding the initial element and writing an updated Tracks element at the end of the file with a seek entry, or updating the existing Tracks element if space permits. While legally possible, putting Tracks at the end is not optimal for playback startup.

Decisions and Action Items

Decisions

The October interim meeting minutes were approved.
The 2026 meeting schedule, including adjustments for IETF overlap and Thanksgiving, was approved.
It was recommended that Errata ID 8615 (Matroska: AttachmentLink MaxAccurs) be classified as "verified" due to it being an editorial error.
It was recommended that Errata ID 8780 (EBML: 0x80 element ID) be classified as "verified" due to its existing usage in Matroska despite being marked reserved in EBML.
The working group has consensus to proceed with the submission of draft-ietf-cellar-matroska-v5.

Action Items

Spencer: Submit the 2026 meeting reservations.
Spencer: Contact AD Ori to discuss the recommended classifications for Matroska Errata ID 8615 and 8616.
Spencer: Coordinate with AD Ori for the formal adoption of draft-ietf-cellar-matroska-v5 as a working group document.
Spencer: Re-familiarize himself with the Matroska v5 build process on GitHub, potentially for a proposal on a different workflow.
Steve: Review the remaining EBML errata (RFC 8797) currently in "reported" status and advise Spencer on their appropriate classification (verified vs. held for document update).
Steve: Reply to the audio-vivid codec registration inquiry, explaining the current limitations and next steps.
Steve: Provide a link to the pull request for timecode source signaling for review by the working group.
Steve: Consider ways to expose Matroska v5 work-in-progress builds (e.g., nightly/weekly HTML/XML) on GitHub to facilitate broader review and engagement.

Next Steps

Monitor the ISG evaluation and telechat ballot for draft-ietf-cellar-tags.
Proceed with the AD consultation and formal submission process for draft-ietf-cellar-matroska-v5.
Continue work on chapter-codex and ffv1-v4 with a target timeline of 2026 for significant progress.
Further classify outstanding errata for EBML (RFC 8797).