Markdown Version | Session Recording

Session Date/Time: 26 Aug 2025 19:00

CELLAR

Summary

The CELLAR Working Group convened to discuss the status of several key documents. Consensus was reached regarding the lack of further restrictions on UTF-8 tag values in cellar-tags, and cellar-codec draft-15 was declared ready for Working Group Last Call. Progress on FFV1v4 remains contingent on GPU implementation feedback, with continued work planned. Discussions around chapter codecs highlighted challenges with specification availability. A key technical discussion revolved around methods for adding contextual information (titles) to Block Additional Mappings, with action items for pull requests outlined. The distinction between errata and update documents was also clarified.

Key Discussion Points

• cellar-tags Document Status:

  • Steve indicated his plan to publish draft-19 of cellar-tags shortly.
  • A significant discussion focused on whether to introduce further restrictions on UTF values within tags beyond basic UTF-8 encoding. Spencer raised concerns about referencing ABNF from an unpublished document and the general principle of restricting tag values.
  • Steve, referencing prior discussions with other participants (not present), conveyed that the proposed changes for draft-19 were acceptable.
  • The sense of those present was that it is too late and undesirable to impose additional restrictions on tag values beyond UTF-8 encoding, as this could break existing Matroska files and goes against established WG practice of not invalidating previously encoded content.
  • Consensus emerged that no further restrictions on tag values beyond UTF-8 encoding would be added to the specification.

• cellar-codec Document Status:

  • Steve confirmed that cellar-codec draft-15 had no outstanding issues or pending merge requests.
  • The document was deemed ready for Working Group Last Call.

• FFV1 Document Status:

  • FFV1 remains in a holding pattern, awaiting further developments from GPU implementations.
  • Jerome stated that efforts are ongoing to update the specification based on feedback from GPU implementations, with a focus on improving parallelization for 8K video support.
  • Michael Neermmyer (not present) has proposed three methods for storing "last bits" without compression, which require benchmarking to determine the optimal approach for speed improvement without significant compression loss.
  • The working group maintains the desire to continue active work on FFV1v4, with an expectation of publication sometime in 2025.

• Chapter Codecs Document Status:

  • The chapter codecs document remains parked with no immediate updates.
  • Jerome noted the continued difficulty in obtaining comprehensive official DVD menu specifications.
  • Experience with other implementations (e.g., VLC's libdvdnav) suggests that relying on widely used, de-facto unofficial specifications might be the only viable path forward.
  • Steve indicated he would have more time to dedicate to this work following the progression of cellar-tags and cellar-codec.

• Block Additional Mappings (Time Code Block Addition):

  • A new pull request from Jerome initiated a discussion on how to add contextual information, such as human-readable titles, to Block Additional Mappings (e.g., to distinguish between different time code types).
  • Options explored included:
    • Using the existing tagging structure (e.g., a Target in Tags) to provide titles, potentially supporting multiple languages.
    • Adding a new Name field directly within the BlockAdditionalMapping structure.
    • Adjusting the definition of the existing BlockAddIDName field.
  • It was clarified that BlockAddIDName currently describes the type of block additional data, not a specific instance's friendly name.
  • Jerome indicated a preference for using tags for flexibility and multi-language support but also acknowledged the utility of a simpler human-friendly name field.
  • Robert expressed concern that not all block additional mappings require additional titling (e.g., HDR data).
  • Steve was amenable to adding elements for version 5 of the specification.

• Errata vs. Update Documents:

  • A clarification was provided on the distinction: An erratum corrects a publication error (grammatical or technical) where the WG's consensus at the time of publication was different from the final text. An update document introduces new content or changes to the specification based on new consensus established after publication. The proposed modification to BlockAddIDName would likely fall under an update rather than an erratum.

Decisions and Action Items

• Decision: No further restrictions beyond UTF-8 encoding will be added to cellar-tags for tag values.
• Action Item: Spencer to initiate the Working Group Last Call (WGLC) for cellar-codec draft-15.
• Action Item: Steve to publish cellar-tags draft-19.
• Action Item: Jerome to submit two separate pull requests on GitHub:
  1. One for adding an ID in the Tags structure to allow for titles for block additions, for consistency and future multi-language support.
  2. One to modify the description of BlockAddIDName to define it as a human-friendly name for block additional data, rather than its type. The nature of this change (erratum or update document) will be further assessed.

Next Steps

• Working Group participants are encouraged to continue reviewing the cellar-codec draft-15 document in preparation for WGLC.
• Jerome's pull requests regarding Block Additional Mappings will be reviewed and discussed on GitHub.
• Further work on FFV1v4 will focus on benchmarking Michael Neermmyer's proposals for storing last bits and integrating GPU implementation feedback.
• The next meeting is scheduled for September 20. The chairs will begin sketching out meeting dates for the first quarter of 2026.