CBOR

Summary

The CBOR Working Group met to discuss the status of active group documents, focusing primarily on a status update and technical refinements for the serialization draft (draft-ietf-cbor-serialization). The group also briefly discussed the last-call status of the Concise Diagnostic Notation (CDN) literals draft (draft-ietf-cbor-edn-literals).

Key Discussion Points

1. Concise Diagnostic Notation (CDN) Literals

Draft: draft-ietf-cbor-edn-literals
Discussion:
- Laurence Lundblade noted that the active mailing list discussions do not require meeting time and represent minor tweaks. There is general consensus on the current content of the document.
- Christian Amsüss indicated that his upcoming review of the draft will be positive and that there are no major issues to discuss.

2. Serialization Draft Status Update and Discussion

Draft: draft-ietf-cbor-serialization
Presentation: Serialization Draft Status Update and Discussion presented by Laurence Lundblade.
Status Details:
- An examples appendix containing 25 highly curated test cases has been added and successfully tested against QCBOR.
- The PR proposing to move the NaN appendix to a wiki was closed. Laurence Lundblade expressed a preference to keep this appendix in the RFC, as NaN-related discussions are ongoing. Carsten Bormann noted that keeping it in the RFC means it requires thorough review by the working group.
- Five pull requests remain open, including sample code for encoding half-precision floats, security considerations (waiting on input), Appendix B location, and clarifications on abstract/Section 1.
- A new draft revision is planned for release in approximately a week.

3. Technical Refinements to draft-ietf-cbor-serialization

General Serialization

Laurence Lundblade proposed reframing "general serialization" to be defined as the complete set of all serializations defined in RFC 8949, with deterministic and preferred serialization treated as subsets or constraints.
Data Model vs. Serialization Constraints: Carsten Bormann pointed out the need to distinguish between serialization constraints and data model constraints. For instance, a constrained device that lacks floating-point support can still implement a compliant generic decoder for the subset of the data model it supports (e.g., if it supports arrays, it must support both definite and indefinite lengths, but it doesn't need to support floats). Laurence Lundblade agreed with this distinction.
Terminology: To avoid mathematical confusion over the word "superset," Michael Richardson suggested using the phrase "complete set," which Laurence Lundblade agreed to adopt.

Special Serialization

Special serialization is used to categorize options outside of "preferred plus" serialization. It currently lists four distinct characteristics:
1. Streaming (indefinite-length encoding)
2. Fixed-width integers
3. Non-trivial NaNs
4. Determinism involving any of the above
Carsten Bormann suggested adding non-preferred floating-point values (analogous to fixed-width integers) to this list, which Laurence Lundblade accepted. Carsten Bormann also suggested checking this list against a previously compiled list of items from Joe Hildebrand.
Вадим Гончаров raised a point regarding library support. Laurence Lundblade clarified that the document explicitly distinguishes between protocol specifications, protocol implementations, and library implementations.

Preferred Plus Serialization

The section on when to use "preferred plus" has been shortened, and the recommendation to use it has been made stronger.
Protocols should use "preferred plus" by default unless they specifically require deterministic serialization or one of the special serializations.
Carsten Bormann clarified that deterministic serialization implies preferred plus, meaning that preferred plus is a necessary but insufficient condition for full determinism. Laurence Lundblade agreed and noted the wording would be adjusted to reflect this.

Decisions and Action Items

Decision: Keep the NaN discussion appendix in draft-ietf-cbor-serialization so it remains part of the final RFC.
Action Item: Laurence Lundblade to integrate comments on the general/special/preferred serialization PR by the end of the week.
Action Item: Laurence Lundblade to cross-check the special serialization categories against Joe Hildebrand's historical list.

Next Steps

Laurence Lundblade plans to publish an updated version of draft-ietf-cbor-serialization in approximately a week, incorporating the half-precision encoding sample, the reframed definitions of general/special/preferred serializations, and minor fixes.
Carsten Bormann will conduct a detailed review of the outstanding PRs upon returning from vacation on Friday, April 5th.

Automatic IETF Minutes