[Agentproto] Re: WG Review: Agent Communication Protocols (agentproto), comment in support with one request

Girish Konda <girish.sai1@gmail.com> Tue, 22 September 2026 07:45 UTC

Received: from mail-qk2-x11.google.com (mail-qk2-x11.google.com [IPv6:2607:f8b0:4864:34::11]) (using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits) key-exchange x25519 server-signature ECDSA (prime256v1) server-digest SHA256) (No client certificate requested) by mx.ietf.org (Postfix) with ESMTPS id 532E631 for <agentproto@ietf.org>; Tue, 22 Sep 2026 07:45:33 +0000 (UTC)
Authentication-Results: mx.ietf.org; dkim=pass header.d=gmail.com header.s=20251104 header.b=d+TvrF2g; arc=pass ("google.com:s=arc-20260327:i=1"); dmarc=pass (policy=none) header.from=gmail.com; spf=pass (mx.ietf.org: domain of girish.sai1@gmail.com designates 2607:f8b0:4864:34::11 as permitted sender) smtp.mailfrom=girish.sai1@gmail.com
Received: by mail-qk2-x11.google.com with SMTP id d75a77b69052e-530de452c33so32313151cf.1 for <agentproto@ietf.org>; Tue, 22 Sep 2026 00:45:33 -0700 (PDT)
ARC-Seal: i=1; a=rsa-sha256; t=1790063127; cv=none; d=google.com; s=arc-20260327; b=Mn27od51foNH/2l0ciZVTHdd94QrQ3XWeRXDG8AlWSGAgLJpNIM3iaKXUfWtgFWpxo XuDvFWflwxKfp780GzqVog6KpQMUWgSJqG1rJCIFwjlFWgecileTusd0U5c2sOE54TIe iSP7euIdaPka7csHk3Fn3QXBCObTkgG8odpz8dTuSBSIDZdBqKn7/y67ou/FzpPDGBLX cm8K/oNIzmBoJRKAC31b52ZhYeetEK8y8sW9FG8UBOiqRdQ6KXN4XNx4pJTcizhtOo1z T519Yebnsy8p8MbwQWTRoH9ChoR0FMab3SHVOXUhtUv6duxw2mdOeyS8GdkSz3vgQio7 +hlQ==
ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20260327; h=cc:to:subject:message-id:date:from:in-reply-to:references :mime-version:dkim-signature; bh=P7JgtXnQ7tBZ1vFx9ivCi3b5gqwilK69ixWKftsFwaQ=; fh=NklVHT0C/oQOpkT7cPUpFXClcRn2/fA9NtIhzNYH4vg=; b=ElWiNr13W+HLnV/7wHRqvwNWuJtAaRGCueHwOyff9hgLOGsK/yQRnpMe8lqCKKS6DX 1yfnyJqe0APYyD+5zcrXLddxeCQtXZLIp5VtK6Qt2GRyoXSQOI49LIzXqX/qIIWA020G FyTFfJ19Xk5/NMdOJ6mVIvVpqPxBuNlhSf0z3gXB+xTx43GomgtONyLSaLE+YNBpmURR Ctthvc/r0PkBHcMiQ6X/6w5kqMpjhFoCrI4y5CWyK+/urvp7liiEhRIQLhzTn3AxLoPc KwK7ghWLyis7xatcVF4iGy5/Dm5o/bZb677O7Gs8NMS1ydJz1LXhTyPljv84VzYFlZMv rTRQ==; darn=ietf.org
ARC-Authentication-Results: i=1; mx.google.com; arc=none
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20251104; t=1790063127; x=1790667927; darn=ietf.org; h=content-type:cc:to:subject:message-id:date:from:in-reply-to :references:mime-version:from:to:cc:subject:date:message-id:reply-to :content-type; bh=P7JgtXnQ7tBZ1vFx9ivCi3b5gqwilK69ixWKftsFwaQ=; b=d+TvrF2gmtf2q03L0LVuSARw4SYV2b7zcuZuDdg0wTOEqu0JoYWEUBz8Q1HUYqXqPL 8WsLVm8NFWT3YLQW0DO/YInLnHhC7iP5gKsJjkkIa1pFiC0zivZ+q/qEknIsIGim6hvf XNhct1905V/VwKTe2/+52MKVJ1EkeBZOL68Sx8USFvBWoW58gEXIzYk3RxaScelCJX7y TYYQttvFMM/NuwjAzokqUv12gCmpkkM9st5OwE+TkzMIj+idUgCbUsDLI3bXr0U0UZOf N9YocVR0+9fsr5hyOAFjwGc5mtbDt6cRerVK99TY7KiisoHpIJxFLDT6eQrN6jHmXTHj fkRg==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20260707; t=1790063127; x=1790667927; h=content-type:cc:to:subject:message-id:date:from:in-reply-to :references:mime-version:x-gm-gg:x-gm-message-state:from:to:cc :subject:date:message-id:reply-to:content-type; bh=P7JgtXnQ7tBZ1vFx9ivCi3b5gqwilK69ixWKftsFwaQ=; b=AfdnR90jlfNo1CrPNltJsiCQO83EflalryYTnlYe83zJQByxdU1ZUmJfghPSNGk7qV ry931QXNHSeVhy5bYxO30oyLO8D4CLSlhqzflr73NpMKn7CuUA0Nkz0XIkiKbGGaad6r RfLDivsuH8IXcxdVAnyWx9rCUzya7vhiCr27hankhhEFqTEPqE/w1yPe1NL3le/DNcs5 yArMNh82k/Bm1oAw1l9IGip5f+Xk25dLg0LEvRGnSXBjh7IaCyag9J2/D86DCJYKBSxL logVC+yxI9n+SxZPbXFdjdSu2CabxmyJiGC7yXvUXAnBhcQWnZkaWHhpQ6Fhna6jX5ej n72A==
X-Forwarded-Encrypted: i=1; AKwUvBwho9HK7p3XW9BJeKcz5ibyJkU1hrubyV5/nErXucGFdvuyysLHfgHAb9VDKBtCqn9Sq05NJYalR+n5@ietf.org
X-Gm-Message-State: AFuF++knZ12hee4Tn8MSE+irpzpLdIdSEHN/+54ugHYAZSaIcghv1Fjd buKB5otZGCwpzsNeGpZx7gv3SP4441d4+Z91G/Bsr/aTn9YO+pOPFOOddQNZOt9TvPNCj8VmdJo VSH1P1T+VAudWsLlkw+yTidJTzQ9Vhdboe4yH
X-Gm-Gg: AYBFou0iOf0ujw4Rh+LOU+lm8+XS1o2hzupZavUAP4kqOJ50WRU4hsVdcjpQM0NDHd8 df3UJZQoxcrLFa0ljhki4B3XunaSST0gcLesE2SM8MFdO4N/EaZfP0qZsE41J20TdnPHaT/hb8c TQdQdAM9810lpHFacop9A7/PafO7eNdudd69u+psgbnUz3wFM4m0k2VAXtOMufcX08erzD5uilI NQUs45tEWtz2bv97dcCx/nmlYt70/ak04IW8vmFOIQMrMJ51cgdAQgjL6xGZamOp7rSZ5soRmLO 7jGdQKby63C9g1CiWlEkjmUAma36QAV9uJxHukMrYZsRm3mL49T7y+8WHKiX5uE=
X-Received: by 2002:ac8:5887:0:b0:530:27a1:88be with SMTP id d75a77b69052e-532d8d9bc68mr44815671cf.30.1790063126851; Tue, 22 Sep 2026 00:45:26 -0700 (PDT)
MIME-Version: 1.0
References: <29283729.644.1790048687028@privateemail.com>
In-Reply-To: <29283729.644.1790048687028@privateemail.com>
From: Girish Konda <girish.sai1@gmail.com>
Date: Tue, 22 Sep 2026 00:44:50 -0700
X-Gm-Features: AcwNN1Uv5WFCq-CUkxRzjpv_QVUy972vIfT1ipZce2Sl7159BLWXtXGR5zLMVUY
Message-ID: <CAMW0qhi8aHuucZjCZ79r0L+koTk7__8qVvdqwooO24iADsJ-0A@mail.gmail.com>
To: Nicholas Templeman <nicholas@csoai.org>
Content-Type: multipart/alternative; boundary="000000000000cd360c065c0d8c1b"
X-Spamd-Bar: --
Message-ID-Hash: S7ZYKULGLOONQZPY4CEH2HKJVMXZNR6P
X-Message-ID-Hash: S7ZYKULGLOONQZPY4CEH2HKJVMXZNR6P
X-MailFrom: girish.sai1@gmail.com
X-Mailman-Rule-Misses: dmarc-mitigation; no-senders; approved; loop; banned-address; emergency; member-moderation; nonmember-moderation; administrivia; implicit-dest; max-recipients; max-size; news-moderation; no-subject; digests; suspicious-header
CC: iesg@ietf.org, agentproto@ietf.org
X-Mailman-Version: 3.3.10
Precedence: list
Subject: [Agentproto] Re: WG Review: Agent Communication Protocols (agentproto), comment in support with one request
List-Id: Agent Communication Protocols <agentproto.ietf.org>
Archived-At: <https://mailarchive.ietf.org/arch/msg/agentproto/TowOwEiBBeHyzW8GP-1xgflapfc>
List-Archive: <https://mailarchive.ietf.org/arch/browse/agentproto>
List-Help: <mailto:agentproto-request@ietf.org?subject=help>
List-Owner: <mailto:agentproto-owner@ietf.org>
List-Post: <mailto:agentproto@ietf.org>
List-Subscribe: <mailto:agentproto-join@ietf.org>
List-Unsubscribe: <mailto:agentproto-leave@ietf.org>

Hey Folks,

I have been following this discussion since the charter went to external
review. I support the charter as scoped and have one comment on the dialog
lifecycle item.

I agree with Sumit Ahuja's comment on 17 September that revocation is the
hard half of the lifecycle, and that a lifecycle which specifies revocation
but assumes a reachable, still-credentialed peer will not cover the cases
Pamela Dingle and Naveed Ihsanullah raised at the Vienna BoF. I would add
one thing to it. Once the peer may be unreachable, the participant that
asked for termination is in one of two states, and they are not the same
state. Either the termination took effect at every hop that was carrying
the dialog, or it was requested and the status at some hops is unknown. The
charter already commits the group to recovery from failures of network
paths and intermediaries and to handling unreachability, so both states are
in scope, but it does not say the protocol has to let a participant tell
them apart.

That matters because the two states call for different behaviour. A
participant that knows termination is confirmed everywhere can release the
context and stop worrying about it. A participant that only knows the
request went out has to keep treating the dialog as possibly live, and has
to pass that uncertainty up to whatever is above it. If the protocol
carries no way to express the difference, implementations will report the
second state as the first, because that is the reporting that looks clean
and nothing on the wire contradicts it.

I want to be clear that this is not a request to reopen audit. PR 120 (
https://github.com/ietf-artarea/charters/pull/120) rules audit data
structures out of scope and I think that is right. Nicholas Templeman made
a related point on 21 September about audit records, that an absent record
should read as not observed rather than as nothing happened, and I am
asking for the same distinction one step earlier, at the moment the
participant is told what happened rather than only in the record that is
written afterwards. If the participant was told the wrong thing then the
record is faithful and still wrong. This sits closer to PR 123 (
https://github.com/ietf-artarea/charters/pull/123) which separates agent
decision-making from protocol behaviour, because what I am describing is
squarely protocol behaviour. It is about what a peer is allowed to say back.

I am not asking for a new deliverable or for new scope. One sentence in the
dialog lifecycle item, saying that a participant must be able to
distinguish termination that is confirmed from termination that is
requested but unconfirmed, would be enough to keep it in front of the group.

Disclosure: I work on multi-tenant control plane infrastructure at
Microsoft, and in my personal time I build agentic and MCP tooling and
participate in a few MCP working groups. Peers that go unreachable and
sessions that may or may not still be alive are a daily problem for me. I
recently built mcp-pacemaker (https://github.com/girishkvs/mcp-pacemaker)
a supervised MCP bridge that keeps servers alive across host restarts,
machine reboots, token expiry and transport failures. This is my personal
opinion and does not reflect my employer's.

Thanks,
Girish Konda

On Mon, Sep 21, 2026 at 8:46 PM Nicholas Templeman <nicholas@csoai.org>
wrote:

> Dear IESG,
>
> I support chartering agentproto as scoped in the 17 September text. The
> scope is
> well bounded: dialog identifiers, lifecycle semantics and context
> propagation,
> with the content exchanged with models expressly outside it, and changes
> to MCP
> and A2A left with their maintainers.
>
> One request on the reference architecture rather than the charter text.
> The key
> considerations name auditability of agent-initiated actions distinct from
> user-initiated actions as a reason for distinct agent identity, and dialog
> identifiers are what a later reader will use to correlate records. It
> would help
> if the reference architecture said plainly that a dialog identifier
> establishes
> correlation and not that an action occurred, so that a record which is
> absent is
> read as not observed rather than as nothing happened. That is a statement
> about
> what the protocol's artefacts can carry, not a new deliverable, and it
> does not
> require naming any effort that is not yet chartered. On that point I agree
> with
> Orie's note of 20 September: coordination with a group that may later
> exist is
> ordinary work with the ADs and does not belong in charter text.
>
> Disclosure: I work on measurement and evidence tooling at CSOAI Ltd. This
> is an
> individual comment, I hold no IPR relevant to the charter, and I am not
> asking
> for anything of ours to be referenced.
>
> Nicholas Templeman
> Council of AI (csoai.org)
> _______________________________________________
> Agentproto mailing list -- agentproto@ietf.org
> To unsubscribe send an email to agentproto-leave@ietf.org
>