The Scope Creep Automated reviews of new AI-related working group submissions

Review · August 21, 2026

DAWN Use Cases

draft-kay-dawn-use-cases -00View on Datatracker →

DAWN Use Cases — draft-kay-dawn-use-cases-00

Problem Statement

This draft catalogs use cases for the Discovery of Agents, Workloads, and Named Entities (DAWN) — the general problem of how one entity (an AI agent, workload, service, or human operator) locates another entity that can satisfy a specific need. The core insight is that in increasingly distributed and agentic computing environments, static configuration is insufficient; entities need a runtime mechanism to discover peers, tools, and resources by capability, policy, jurisdiction, or resource type.

How the Spec Fits In

This is a use-case and requirements document, not a protocol specification. It belongs to the emerging DAWN family of drafts (alongside terminology, architecture, and more specific mechanism drafts) that are working toward a discovery framework for agentic and distributed systems. It has no dependency on OAuth or related identity RFCs. It is more comparable to informational requirements documents in the networking and routing area (PCE, CATS, etc.) than to protocol specs. This draft explicitly defers all protocol decisions — registration, selection, negotiation, and communication — to other drafts, positioning itself purely as a problem-space survey.

High-Level Summary

The document defines four categories of discovery:

  • Capability-Oriented Discovery — finding an entity that can perform a specific function (e.g., an agentic workflow decomposing a task and finding the right sub-agent for each step).
  • Resource-Oriented Discovery — locating datasets, compute resources, models, or inference services by type, availability, jurisdiction, or policy.
  • Administrative Scope Extensions — cross-organizational discovery where one party's agent needs to find an authorized agent or service at another organization, while both parties retain control over what they expose.
  • Operational Discovery — discovery used for audit, incident response, compliance, and management by human operators or AI-assisted operations systems.

The document also identifies five entity classes that discovery applies to (AI agents, software services, compute workloads, network functions, and application endpoints), though these sections are all marked TBD in this revision.

A notable architectural principle: the document explicitly states that a single central registry should not be required — federated and brokered discovery is a first-class goal.

Interoperability

As a use-case document, this draft is not meant to directly produce interoperable implementations. Evaluated on its own terms — whether it adequately motivates and frames a future protocol effort — it does a reasonable job of defining the problem space, though there are gaps.

Concerns about completeness: The five entity-class sections (Section 7) are all marked TBD in this -00 draft. These sections are described as the structural core of the use case catalog, so their absence means the document is substantially incomplete. The four discovery-category sections (Section 6) provide useful illustrative examples, but without Section 7 being fleshed out, it is unclear whether the four categories cover the full range of behavior needed across all entity types.

Overlap risk: Several of the examples overlap significantly with existing work. The cross-domain administrative discovery examples (Section 6.3) closely resemble problems already addressed in federation and trust frameworks in identity standards. The document does not reference this prior art, which could lead to the DAWN effort reinventing mechanisms already standardized elsewhere.

Could be removed: The "Discovery Pattern" section (Section 5) is generic to the point of being obvious — any lookup mechanism follows the described five steps. It adds length without contributing meaningful requirements. Similarly, the Operational Considerations section largely restates observations already implied by the use case descriptions.

Writing Quality

The writing is generally clear and uses appropriate IETF informational document style. Sentences are well-formed and grammatically correct. The structure is logical and the scope limitations are stated clearly up front.

The draft is at an appropriate level of abstraction for a -00 use-case document. However, with Section 7 entirely TBD, the draft reads as a work-in-progress outline rather than a complete document. Publishing a -00 with major sections blank is not unusual in IETF, but it makes evaluation of the overall framing difficult. Reviewers should treat this as a preliminary structure proposal more than a substantive specification.

The acknowledgments (Adrian Farrel, Jim Mozley) suggest the draft has had at least informal review from experienced IETF participants before submission.

Background of the Authors

Daniel King ([email protected], Old Dog Consulting) — Highly experienced IETF contributor with 22 published RFCs spanning PCE, MPLS, GMPLS, and YANG, dating back to 2006. He currently has 17 active drafts, indicating deep ongoing involvement. His participation in DAWN provides the group with significant process credibility.

Kehan Yao ([email protected], China Mobile) — No published RFCs, but has 14 active drafts and 41 expired drafts, suggesting sustained IETF participation across CATS, AI networking, and network management areas without having completed the publication cycle yet. A productive participant but not yet a proven RFC author.

Ken Adler ([email protected], Indeed) — No published RFCs and only this single active draft. Appears to be a newcomer to IETF contribution, likely bringing practitioner perspective from industry rather than established standards background.

The author mix is notable: one experienced network-area veteran (King), one China Mobile participant with broad but unpublished drafts activity (Yao), and one apparent industry newcomer (Adler). This is a common pattern for emerging-technology work items where practitioners are driving the problem statement.