[Agentproto] Re: [iesg] Roman Danyliw's No Objection on charter-ietf-agentproto-00-02: (with COMMENT)

["Charles Eckel (eckelcu)" <[e-mail]>]

Hi Roman,

There is work in progress on the charter in GitHub, including some changes that hopefully address your remaining concerns. Please see inline.

On Sep 17, 2026, at 2:00 AM, Charles Eckel (eckelcu) <[e-mail]> wrote:

Hi Roman,

Thanks for reviewing and providing these comments. Please see inline.

On Sep 16, 2026, at 8:17 PM, Roman Danyliw via Datatracker <[e-mail]> wrote:

Roman Danyliw has entered the following ballot position for
charter-ietf-agentproto-00-02: No Objection

When responding, please keep the subject line intact and reply to all
email addresses included in the To and CC lines. (Feel free to cut this
introductory paragraph, however.)



The document, along with other ballot positions, can be found here:
https://datatracker.ietf.org/doc/charter-ietf-agentproto/



----------------------------------------------------------------------
COMMENT:
----------------------------------------------------------------------

** “An AI agent is an autonomous, adaptive software system that uses AI models
to complete a specific task on behalf of a human user, another AI agent, or an
invoking system.  AI agents interact with users and other agents …”

-- Consider refining or clarifying “invoking system” to make the distinction
clear.  Isn’t an “AI agent” an “invoking system” too?

I believe it is, but my understanding is that “invoking system” is not limited to AI agents.


-- In the next sentence, AI agents are said to interact with “users” and “other
agents”.  What happened to “invoking systems”?

My assumption is that interaction with invoking systems would not typically be via multiple modalities. If it is, then yes, it could be added.


** “An AI agent is an autonomous, adaptive software system that uses AI models
to complete a specific task …” vs. “… and are capable of independent
decision-making, tool invocation, and task completion”

What’s the difference in these definitions of an AI agent?  “Autonomous”
already implies “independent”.  “Complete a specific task” already implies
“task completion”.

They are complementary. Tool invocation is implied in the first, but called out explicitly in the second. These definitions could be merged if helpful.


** “User-to-agent, agent-to-agent, and agent-to-tool …”

-- This taxonomy seems to exclude the “invoking system” component identified in
the first sentence.

-- The narrative text prior doesn’t define “tools” in a way that distinguishes
them from agents.

We have an open issue to add a definition of tool:
https://github.com/ietf-artarea/charters/issues/107

Would it be helpful if we added an issue to define [or remove] invoking system?


** “The dialog context consists of the identifiers and lifecycle state needed
to correlate and maintain an agentic dialog …”

-- Dialog would benefit from definition to understand how low in stack this
charter intends to operate.  Med asked a related question on layer.

-- This text does little to clarify terms as it turns the term “dialog”
previously used as a noun into an adjective, “dialog _context_”.

I added a new issue specifically for a definition of dialog,
https://github.com/ietf-artarea/charters/issues/114

The following definition:

"A dialog is the correlated sequence of interactions between two or more participants over the course of a task.”,

is added in this PR: https://github.com/ietf-artarea/charters/pull/125


These is a related issue to address layering based on comments from Mahesh:
https://github.com/ietf-artarea/charters/issues/112

The following PR is mean to address this: https://github.com/ietf-artarea/charters/pull/124



** “It does not include the content …”

Is this text clarifying what is “dialog” or “dialog context”?

The dialog context.


** “… when a participant changes device or network attachment, or when a dialog
resumes after an interruption.”

Is a “participant” a user, agent, tool, or invoking system?  Intermediary?

I believe this refers to a participant in the dialog, which is a human user, AI agent, or tool.


** “The working group will analyze the privacy implications of dialog
correlation and specify mitigations.

Because dialog identifiers may persist across intermediaries and trust
boundaries, the working group will analyze the privacy implications of dialog
correlation and specify mitigations.”

-- What is the difference with implied scope or tasking between these two
paragraphs?  They both read like the same work except the second paragraph
provides motivation.

This seems to be the result of a botched merge on my part. The second version names the specific property creating the risk. I will keep it and remove the first.


-- Are the threats due to correlation here that a single dialog context is
being shared around trust boundaries; or that one could correlate multiple
dialog contexts?  “Correlation” is being used a few times in this charter.

Both.

We merged PR #122 to try to clarify this.
https://github.com/ietf-artarea/charters/pull/122



** "Standardization of agent behavior, decision-making, or planning semantics."

The vague "standardization of _agent behavior_" needs to be reconciled with
this WG in fact changing agent behavior by enabling it to emit and use "Agentic
Dialog Management Protocol" (assume adoption of the work).

Good point. My initial thought is that this WG will define how an agent MUST “behave” in order to support the Agentic Dialog Management Protocol, but an agent might not necessarily support the Agentic Dialog Management Protocol, and its behavior will almost certainly go far beyond anything related to support of the Agentic Dialog Management Protocol.

We merged PR #123 to clarify this:
https://github.com/ietf-artarea/charters/pull/123

Please let us know if this addresses your concerns or if additional change is still needed.
Thanks for helping us improve the charter.

Cheers,
Charles


Cheers,
Charles





_______________________________________________
iesg mailing list -- [e-mail]
To unsubscribe send an email to [e-mail]