[dmsc] Re: Fwd: New Version Notification for draft-feng-agentproto-idl-info-model-00.txt

[Chris Hood <[e-mail]>]

For your reference, review AGTP-API which places semantics inside of the methods on a dedicated transport. Benchmark tests show up to 29% improvement in accuracy for agent transaction. So the argument that agents need more than transport, endpoints, etc, is valid unless it's packaged as a first class citizen on the transfer layer. https://www.ietf.org/archive/id/draft-hood-agtp-apiChris
-------- Original message --------From: chong feng <[e-mail]> Date: 7/27/26  4:47 PM  (GMT-08:00) To: [e-mail] Cc: [e-mail], [e-mail], [e-mail] Subject: [dmsc] Fwd: New Version Notification for draft-feng-agentproto-idl-info-model-00.txt Hi all,I would like to share a new individual draft for review and discussion:  IDL: A Semantic Information Model for Agent Communication  draft-feng-agentproto-idl-info-model-00  https://datatracker.ietf.org/doc/draft-feng-agentproto-idl-info-model/IDL (Intent Description Language) is a protocol-independent semanticinformation model for communication among autonomous agents and betweenagents and tools.The motivation is that agent communication needs more than transport,message framing, endpoint discovery, or tool schemas. Before agents cancollaborate, they need a shared way to describe questions such as:* Who is the participating agent, and how is its identity bound to an  authenticated principal?* What capabilities can it contribute to an intent?* Within what declared autonomy and authority boundaries may it act?* Which intent lifecycle roles and interaction patterns does it support?* What context does it require, maintain, or share?* Which observability, control, and intervention interfaces does it expose?The draft therefore defines an information model covering agent identity,capability declarations, autonomy envelopes, intent participation, context,and governance interfaces.IDL is intended to complement agent communication and discovery protocols,rather than define another transport or messaging protocol. It also doesnot standardize agent behavior, decision-making, planning, or reasoningalgorithms. The JSON representation and interaction sequences in the draftare currently illustrative.I believe this semantic layer may be relevant to several related discussions:* agentproto, for capability exchange, agent selection, context transfer,  negotiation, and protocol-visible interaction semantics;* DAWN, for describing the semantics of discovered agents and workloads;* DMSC, for secure multi-agent collaboration and gateway-mediated  interactions; and* AINETOPS, for autonomy, governance, and operational control of agents.Comments and suggestions would be very welcome, especially on:* whether this information-model layer fits the agentproto problem space;* overlap or alignment with existing agent description mechanisms;* the identity, credential-binding, and autonomy models;* missing semantics needed for agent selection and collaboration; and* how this model could be mapped to discovery or communication protocols.Best regards,Chong Feng---------- Forwarded message ---------发件人： <[e-mail]>Date: 2026年7月28日周二 07:40Subject: New Version Notification fordraft-feng-agentproto-idl-info-model-00.txtTo: Chong Feng <[e-mail]>A new version of Internet-Draft draft-feng-agentproto-idl-info-model-00.txthas been successfully submitted by Chong Feng and posted to theIETF repository.Name:     draft-feng-agentproto-idl-info-modelRevision: 00Title:    IDL: A Semantic Information Model for Agent CommunicationDate:     2026-07-28Group:    Individual SubmissionPages:    41URL:https://www.ietf.org/archive/id/draft-feng-agentproto-idl-info-model-00.txtStatus:https://datatracker.ietf.org/doc/draft-feng-agentproto-idl-info-model/HTMLized:https://datatracker.ietf.org/doc/html/draft-feng-agentproto-idl-info-modelAbstract:   This document defines the information model for IDL (Intent   Description Language), a protocol-independent semantic model for   communication among autonomous AI agents and between agents and   tools.   Autonomous agents are fundamentally different.  An agent possesses   its own reasoning, may negotiate parameters, may refuse or modify   requests, and may initiate interactions on its own behalf.   Describing an agent as if it were a deterministic endpoint loses   exactly the properties that make it an agent.   IDL answers: "what can this autonomous participant contribute to an   intent, under what constraints, and through what interaction   patterns?"  IDL is not a transport, session, discovery, or messaging   protocol.  It is a semantic model that enables heterogeneous agents   to describe identity, capabilities, autonomy boundaries, context   requirements, governance interfaces, and intent participation before   and during collaboration.   This document defines the IDL information model and its relationship   to agent communication protocols and governance frameworks.  It does   not define a normative JSON representation or a serialization-level   conformance language.  Protocol-specific mappings, including an AIN/   CDL profile, are informative.The IETF SecretariatDMSC mailing listTo subscribe or manage your subscription: https://mailman3.ietf.org/mailman3/lists/dmsc.ietf.org/