CBOR

IETF CBOR Working Group Interim Meeting Date: 2026-07 (Interim)

Summary

The session focused primarily on resolving outstanding issues for draft-ietf-cbor-edn-literals. The discussion highlighted a fundamental architectural debate regarding whether the ABNF for CBOR Extended Diagnostic Notation (EDN) should follow a monolithic/integrated model or an isolated/extension-based model. Other topics included deterministic encoding indicators and the inclusion of Constrained Resource Identifiers (CRI). Due to the length of the EDN discussion, the presentation on "Partial CBOR implementations" was deferred to the mailing list.

Key Discussion Points

EDN Literals (draft-ietf-cbor-edn-literals)

Carsten Bormann presented the status of several issues following the publication of version -22 of the draft (EDN Slides).

Issue 82 (C-style comments): Included in version -22. Carsten Bormann noted that while not everyone is perfectly satisfied, there is a convergence.
Padding Tolerance: Discussion centered on whether to allow lenient padding for base64. Carsten Bormann suggested sticking strictly to RFC 4648 (Section 4 and 5), requiring either correct padding or no padding.
Issue 86 (Float and Same EDN extensions): These will remain in a separate draft to avoid controversy in the base document.

Integrated vs. Isolated ABNF Model

A significant portion of the meeting was dedicated to the extensibility model of the EDN grammar (Issues 88, 91, and PR 92).

Rohan Mahy argued for an integrated/monolithic ABNF. He contended that a single BNF is easier for implementers to build fast, single-pass parsers and suggested that extension writers should bear the burden of ensuring their grammar integrates cleanly. He demonstrated an integrated ABNF in a recent PR.
Carsten Bormann advocated for the "bazaar" or isolated model, arguing that it allows extensions to develop in parallel without requiring the main ABNF to be recompiled or edited for every new addition. He emphasized that EDN is a diagnostic notation, not a high-performance interchange format, and architectural isolation is more robust.
Christian Amsüss expressed concern that requiring extension writers to provide integrated ABNF would be difficult for non-experts and might lead to broken or ambiguous grammars if they simply "click together" existing grammars like URI specs.
Rohan Mahy countered that a two-pass parser model shifts work to the implementers and complicates the development of tooling like code coloring plugins.

Encoding Indicators and Determinism

The group discussed expressing deterministic encoding constraints within EDN, potentially utilizing the draft-ietf-cbor-serialization context.

Carsten Bormann proposed using encoding indicators (e.g., _C for CDE) to specify serialization constraints without changing the data model.
Вадим Гончаров (Vadim Goncharov) raised questions regarding the namespace for these indicators and whether they could eventually take parameters.
Rohan Mahy noted it is useful for diagnostics to know if a binary string was originally encoded as "preferred" or not.

Constrained Resource Identifiers (CRI)

Marco Tiloca and Christian Amsüss both supported keeping the definition of CRI within draft-ietf-cbor-edn-literals, noting its active use in other working groups like CORE and ACE.

Raw Strings (Backticks)

Carsten Bormann confirmed the backtick syntax was derived from a survey of 10 modern programming languages to ensure raw string handling is robust.
Вадим Гончаров (Vadim Goncharov) supported the backtick mechanism as a way to include complex data, such as entire programming languages, without "escaping mess."

Partial CBOR Implementations

The presentation on "Partial CBOR implementations" by Laurence Lundblade was deferred due to time constraints.

Decisions and Action Items

Decision: Issue 82 (C-style comments) will be closed based on the current implementation in version -22.
Decision: Issue 86 (incorporating float/same) will be closed as "won't fix" for the current document; these will remain in a separate specification.
Action Item: Barry Leiba will initiate a thread on the mailing list to gauge consensus on the ABNF model (Integrated vs. Isolated) to break the deadlock between the two proposed approaches.

Next Steps

The chairs will evaluate the mailing list discussion on ABNF to determine the path forward for PR 92.
Discussion on encoding indicators for deterministic serialization will continue on the mailing list.
The deferred presentation on partial implementations will be rescheduled or handled via the list.

Automatic IETF Minutes