CORE

CoRE Working Group Interim Meeting Date: 2026-06-11 (Interim 06) Chairs: Marco Tiloca, Carsten Bormann

Summary

The CoRE working group held an interim meeting to discuss optimization of join proxy discovery (relevant to ANIMA), resource type flexibility in draft-ietf-core-coap-pubsub, and the technical mapping of SIDs to Normalized Instance Trees (NITs) in the context of CORECONF and YANG-CBOR.

Key Discussion Points

Join Proxy Discovery and Payload Shortening

Esko Dijk presented two approaches to shorten the payload of discovery responses for constrained join proxies (connecting to work in ANIMA). The goal is to reduce the size of link-format responses sent over link-local multicast.

Approach 1: URI Reference with Empty Hostname
- Proposed format: <//:port/path>;rt="core.jp". This uses a relative-ref with a zero-length registered-name for the host.
- Michael Richardson noted that // is used in browsers to maintain schemes across different authorities.
- Carsten Bormann cautioned that while URI references are standard, an empty hostname with a port number should be checked against RFC 3986 resolution rules to ensure it does not cause implementation conflicts.
- Marco Tiloca and Maria Matějka discussed the semantic implications of the scheme (CoAP vs. CoAPS). The resource type core.jp could implicitly signal the requirement for security (DTLS).
Approach 2: Custom Attribute
- Proposed format: <>;rt="core.jp";jp="port". This moves the port information to a new attribute to keep the URI target empty.
- Carsten Bormann noted that an empty target <> refers to the resource where the request was made (e.g., .well-known/core), which might be safer from a BCP 190 perspective than pointing to the root.

CoAP Pubsub (draft-ietf-core-coap-pubsub)

Esko Dijk raised a question regarding the core.ps.data resource type. Currently, it is a default, but the specification was unclear on how extensions or future versions should be handled.

Discussion: The group discussed whether a resource should support multiple resource types (RTs).
Carsten Bormann emphasized that the protocol must allow for smooth transitions and evolution. A data resource should be able to offer both the standard core.ps.data and a potential future version simultaneously.
Outcome: The group favored allowing multiple space-separated resource types in the configuration, effectively treating the RT field as an array or a space-separated string to ensure future-proofing and extensibility.

YANG NITs and SIDs

The group discussed a recent mailing list thread regarding how Actions and RPCs are represented in Normalized Instance Trees (NITs).

Vojtěch Vilímek argued that since Actions and RPCs are not "data nodes" in the YANG sense, they should not be included in instance identifier paths.
Carsten Bormann explained that in the context of YANG-CBOR and SIDs, the goal is maximum compression. Including the action point and its input/output details in the NIT allows for a single SID to cover the entire path.
Discussion: There is a technical distinction between the schema path (used for NITs) and the instance identifier path. Carsten Bormann argued that a SID must map to a single NIT (a functional mapping), even if the string representation is unique to the YANG-CBOR context.

Decisions and Action Items

Pubsub: Esko Dijk and the authors of draft-ietf-core-coap-pubsub will add text to clarify that core.ps.data is an extension point and that multiple resource types can be associated with a topic.
Join Proxy: Esko Dijk will review RFC 3986 resolution rules regarding empty hostnames and provide a proposal for the join proxy discovery format to the CoRE and ANIMA mailing lists.

Next Steps

Continued discussion on the mailing list regarding the terminology of NITs and SIDs to resolve the discrepancies between schema paths and instance identifiers.
The chairs will schedule the next interim in two weeks.

Presentation Materials

Chairs' slides

Automatic IETF Minutes