Session Date/Time: 04 Feb 2026 15:00

CBOR

Summary

The CBOR Working Group meeting addressed the status of draft-ietf-cbor-serializer-info, focusing on proposed recommendations for serialization practices in various protocol contexts and libraries. Key discussions included defining "framework" versus "end-to-end" protocols, recommendations for serialization choices, potential naming for "ordinary serialization," and a consensus to prohibit future CBOR tag definitions from modifying the CBOR data model. The meeting experienced a delayed start due to chair absence but proceeded once Paul Hoffman joined.

Key Discussion Points

Meeting Start and Chair Absence: The meeting began late due to the absence of the chairs, Barry Leiba and Paul Hoffman. After some delay and attempts to contact them, Paul Hoffman joined approximately 12 minutes into the scheduled time, apologizing for the mix-up in scheduling.
draft-ietf-cbor-serializer-info Status Update (Lawrence Lundy):
- Lawrence provided an update on the 02 version of the draft.
- New Sections: Added "Recommendations," "When to Use General Serialization," and "Special Serializations."
- New Appendices:
  - "Big Numbers in the Data Model" (explaining how tags 2 and 3 modify the data model).
  - "How to Implement Big Numbers."
  - An example about serialization in COSE.
- Decoding of Big Numbers: The rule for decoding leading zeros and empty strings for big numbers was changed from "should" to "must" on the decode side, representing a small divergence from RFC 8949 for interoperability. Karsten M. Ivens raised a question about this change, noting that preferred serialization (which doesn't have these cases) is being implicitly handled. Lawrence suggested further discussion on the mailing list if needed.
- Open Issues & PRs: 8 open issues remain, with the top action item being where to state the prohibition against future tags modifying the data model. 2 open PRs, including a general review from Rohan and Anders Fogh's comments on the deterministic serialization section, are pending.
Recommendations Section Discussion (draft-ietf-cbor-serializer-info):
- Protocol Types (Framework vs. End-to-End): Lawrence proposed a distinction:
  - Framework Protocols (e.g., CWT, COSE): Many options, no negotiation, don't guarantee interoperability without additional help.
  - End-to-End Protocols (e.g., FIDO, TLS): Must guarantee interoperability between conforming implementations.
- Serialization Guidelines for Protocol Specifications:
  - Framework Protocols: Initially recommended saying "nothing" about serialization, letting the deploying environment decide, except in rare deterministic cases.
  - End-to-End Protocols: Must specify serialization, recommending "ordinary serialization" or "deterministic" (as a stand-in for ordinary).
- Serialization Guidelines for Libraries:
  - CBOR Libraries: Should use ordinary serialization as a default for encoding, must be able to decode ordinary, and should support deterministic.
  - Framework Protocol Libraries: Suggested to use ordinary serialization, may use deterministic.
  - End-to-End Protocol Libraries: Should adhere to the protocol document's specification.
- Feedback on Recommendations:
  - Rohan and Karsten M. Ivens suggested that framework protocols can say something about serialization, especially if they have specific media types or deployment contexts. Karsten noted that framework protocols are in a better position to guide end-to-end protocols.
  - There was a sense that the draft should relax the strict "say nothing" for framework protocols, allowing them to be specific if their use case warrants it.
  - Karsten also pointed out that "ordinary serialization" is not currently a default in existing protocols and urged clarity on whether recommendations apply to encoders or decoders.
- Consensus on Framing: Lawrence emphasized the need for strong consensus on the proposed framing of framework/end-to-end protocols and libraries. Paul Hoffman suggested soliciting rough consensus on parts of the document (e.g., the breakdown of protocol types) via the mailing list.
Naming Discussion for "Ordinary Serialization":
- Lawrence's intent for "ordinary" was to convey an "everyday, broad use, easy choice." He was open to other names.
- Karsten M. Ivens proposed "mainstream serialization," arguing it better captures the intent to "re-steer" the CBOR community towards specific guidance.
- "Basic" was discussed and rejected as it might imply the simplest implementation, which "ordinary serialization" is not. The group seemed to lean towards "mainstream."
Prohibition of Future Tags Modifying the Data Model:
- There was consensus that future tag definitions are not allowed to modify the CBOR data model in the way tags 2 and 3 do (which are grandfathered).
- The discussion revolved around where this normative statement should be placed. Options included the serializer-info document, a separate data model draft, or being embedded in IANA registry rules.
- Paul Hoffman (co-chair), after consulting with the AD, advised Lawrence to place the normative text in the serializer-info document for now. The chairs and ADs would later determine the final placement (e.g., main body, separate document, IANA).
- Lawrence indicated he would add the text and might experiment with a separate data model draft to see if it provides more clarity.
- Karsten M. Ivens noted that existing "unliked" tags are still registered but new tags with similar properties are not recommended.

Decisions and Action Items

Lawrence Lundy (Editor):
- Revise the "Recommendations" section of draft-ietf-cbor-serializer-info to allow framework protocols to specify serialization if their context requires it, rather than strictly "saying nothing."
- Consider "mainstream serialization" as a replacement for "ordinary serialization."
- Add normative text prohibiting future CBOR tags from modifying the data model to draft-ietf-cbor-serializer-info. He may experiment with a separate data model draft to explore this further.
- Clarify language around "use" of serialization in the document to specify "encoder" or "decoder" constraints.
- Seek rough consensus on the mailing list for the framework/end-to-end/libraries breakdown in the recommendations section.
Paul Hoffman (Chair):
- Open an issue regarding the overall organization of the document and the use of appendices versus the main body.

Next Steps

Continued discussion on the mailing list for pending issues and PRs.
The chairs will consult with the AD regarding the final placement of the normative text prohibiting future tags from modifying the data model.
A decision on scheduling the next meeting will be made on the mailing list.