[OAUTH-WG] Re: 回复:Re: Call for interest: fine-grained authorization in OAuth (IETF 126 follow-up)
Ron Bartor <roniamdev@gmail.com> Fri, 02 October 2026 19:19 UTC
Received: from mail-ej2-x0d.google.com (mail-ej2-x0d.google.com [IPv6:2a00:1450:4864:34::d]) (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 25BEF42 for <oauth@ietf.org>; Fri, 02 Oct 2026 19:19:48 +0000 (UTC)
Authentication-Results: mx.ietf.org; dkim=pass header.d=gmail.com header.s=20251104 header.b="L82/3Xdi"; spf=pass (mx.ietf.org: domain of roniamdev@gmail.com designates 2a00:1450:4864:34::d as permitted sender) smtp.mailfrom=roniamdev@gmail.com; dmarc=pass (policy=none) header.from=gmail.com; arc=pass ("google.com:s=arc-20260327:i=1")
Received: by mail-ej2-x0d.google.com with SMTP id a640c23a62f3a-c2e0b486d91so37405166b.2 for <oauth@ietf.org>; Fri, 02 Oct 2026 12:19:48 -0700 (PDT)
ARC-Seal: i=1; a=rsa-sha256; t=1790968787; cv=none; d=google.com; s=arc-20260327; b=ju2OObMKgLV4oKCisQ7kt+lh/+I7g98gi7RzFMGUcXfaADFnFDAA9pXvDB26LlpP2W 1Lo1yeUBBhqKb5v8vcriAoFsJgcWQGFSySBeslR2JA6OaPqrbvK/EBGlZpXDXX6WVCty +FoT4Qn2lPGFughWpwsauH+qAVWoa/DiDdZ8FMZ/CMzeHBizqOg8dy1RpdsbXmEbz6/K x/ht541F+w+DicsJx38KKFM9PUvZiWqOCC3p8vYoPHFZ8hVHvaf+lG3eGqTd/BnY+5Nl ttc/+0bnnZV3GNLzQexWITeMwy8Pv28GidGuy5z+kUOHj9nWGJOSsnrqV+mdg3vCQrOX Jfgw==
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=vBRSFOTYlUVNpQeHczNFCUOLmIzgAF/MIn+Bir4KzWM=; fh=SqYQuy5Vb76PDwBNUFXw3DM4sz6hz+xH2MKdKM5H9U4=; b=JBiVe+9tawgCXvedFKiRB2G7OW5ks3Iuj5l4q7V9P2UbpMDTcdZZBisUVzLSYqDnbc 7JXfPbMjFpJgTNSEhwC/paVcm0zi/ntEjzuy5ZMd+R9PzuXisN3u/M1HBEGtGXGHeJn+ qAKd0iGUDDN68geMDYyo8Ckjo/jieNRrgnyj7a/FmooeuI0+1ThBg47LrAGQ1z1FuEUX 4fv6TAp5q5nYmkc4ZSc46ppH6/6yG2lbq64DGXAY5hGweHY73XpAz5lm+ZgKJDrNhV4c uhM7RydN3Pi4Lcm+FKIUbijFjtE8QeHdpUlzBnaqVtpKCP4uMrbfG8MH7OUp/369ixh/ Xmug==; 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=1790968787; x=1791573587; 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=vBRSFOTYlUVNpQeHczNFCUOLmIzgAF/MIn+Bir4KzWM=; b=L82/3XdiG3nFGTjPMOvEn/a5uRbAEv3x7ICC8BrlRAEoBbremrotAYjXzoga/ouw2k cMnjnGRqY1LYcWGdMkERcKkpHU4Iae/5S2IQ6HMtulBfxfg7W3/aYspyjh4H3pHPVWWS YEmwiQCd52cbxhdFJXhhwFDabp5ST+L6NXD6/OszJDD5uzeEOO5oSVgm6OEC2ESz1tpm pCHObZarKcVusMw8QK+e3QBnxQjVF+oqo/q7MFaH5W9eLgH+Ahb/FrPAyB3JiSktp9Tc ZqdDEMsIZGrmhF5edv0qdek5MyGM/D60D9kF3MU5oHcfO8oGKiNAbTi4d7OwYOSpaq5Q 7J6g==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20260707; t=1790968787; x=1791573587; 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=vBRSFOTYlUVNpQeHczNFCUOLmIzgAF/MIn+Bir4KzWM=; b=b+5RCF4UUbW7EYABlIPXgM1MngClkrNdIJx3el62yKu+fNlSJdto9Z7yLCYeI2+0XL K4t2Rr9YNURgKSFVhmZrDL6yeWY7JVzunz35uOBpdfQMA/T2zd/jqnHzECFj6R6Ecp79 +SdfEQFLTCBNjOxRACJNPYgXDBciXW8eKoQbzyUUoFPC93ETUr6DHhDE95dV+LHnCMXi sEDKULuZ/Qj18a9XYwdCbieDAEIh6P0kCmLtaOLh1jXaEmdwcAyHKBNtas7haVmruZ5H fhPAYVoMSHroaKkx4lwYHm0T6hX6S5H705ePEu+1yMhLziOc/e4I7VN1VNi2iF0rwTsK bdrA==
X-Forwarded-Encrypted: i=1; AKwUvBwpHZdtqN0G5V0y4lbeeo0GMCC/ig0WyyqxS7HeVZ+BR9qGxqdV+LrgYTSeJDatSdUwSe5yLw==@ietf.org
X-Gm-Message-State: AFuF++kzB8Td2m4yztEZ53wk3oEaMwj8lEgyyj0YiypjTqEk6VOw87jw HENv5sY/D5TzLVCOijSOCMEUSTiAST54AXxEe1BE5ojtiZYNzZyq5DXKZbJhsxAyYcsVmvUqyDX 2kklgEONf3csPvSNoNjIJL7XnFUNVM+3+vvr0MbA=
X-Gm-Gg: AYBFou333Qx92TzmpfjfJ1x+DHORml6R1g1zYy7kYnSUJ209UJV6gt2ygEe/to1t14o eox0wOUy3NI5DGiyFXIObwTk3dHJxecXDzUgpS1qm+RmMljhb4gw+GDd2eWtsiFMH9klMm+bfpX Ent1InA9czpBONEfGKkI66E8HXTZHACcdtl9sbMWlBoaiAuc+XVaydPa+T6SVlZHNKEUxz7TTQo KZt/npIeOv+nby/LqLvmU4U0hmsqojBlDP0TPIDwo41NLGsxefKOPKmYwoS400Fk2ObEf5Z3LSm b5h/Nv4cviVv/ZWlNx8rQ6JPRZamtRVK+KIwIDFuJmJDzDyJIYDb5ALy
X-Received: by 2002:a17:907:3d56:b0:c29:d1b2:35df with SMTP id a640c23a62f3a-c2e4aac026bmr319562066b.10.1790968786264; Fri, 02 Oct 2026 12:19:46 -0700 (PDT)
MIME-Version: 1.0
References: <db89ae6c-f90e-4fee-add3-0c66a7746053.max.ldp@alibaba-inc.com> <CAOfgHgo7qVMu0iGAdsyQySs91YrcOpcnLXfGtV_maSW8ncV56Q@mail.gmail.com> <CANr0dJFS98Wvum5MERgS_TLx07XTjNx6e8vuxLkEexTi_Q_HLA@mail.gmail.com> <PH0PR18MB46859723EAEE7FF14514E822D98B2@PH0PR18MB4685.namprd18.prod.outlook.com> <MMZPR04MB481271FA9666C306914202FFBD938B2@MMZPR04MB481271.eurprd04.prod.outlook.com> <CANr0dJEOSqvqqKOO+URUdg9Lf1qDHYib+LCM1b=5BJfqTZ+Fsw@mail.gmail.com> <MMZPR04MB4812714E1A971557FEE59DEF1A938A2@MMZPR04MB481271.eurprd04.prod.outlook.com> <PH0PR18MB468572ECC9F892BD917BD221D98A2@PH0PR18MB4685.namprd18.prod.outlook.com> <dbe4a573-9d3e-4a76-802d-22d39d177e26.max.ldp@alibaba-inc.com> <YQBPR0101MB471530E51122A86E5C535811F98A2@YQBPR0101MB4715.CANPRD01.PROD.OUTLOOK.COM> <PH0PR18MB46859800A6AE443CFFA1F9F1D98A2@PH0PR18MB4685.namprd18.prod.outlook.com> <YQBPR0101MB4715835AC5BE6BD874F64BCAF9892@YQBPR0101MB4715.CANPRD01.PROD.OUTLOOK.COM>
In-Reply-To: <YQBPR0101MB4715835AC5BE6BD874F64BCAF9892@YQBPR0101MB4715.CANPRD01.PROD.OUTLOOK.COM>
From: Ron Bartor <roniamdev@gmail.com>
Date: Fri, 02 Oct 2026 12:19:33 -0700
X-Gm-Features: AclHuK_Vu_h8PkBIWlvrMRfof4VUCvA3gRe7DTUkhpFzFuALpcMKgbYHBXw6Kws
Message-ID: <CANr0dJF0uyuX5QxcRtaG7ez8dtZsmb9i5XKsZnk297EGUwrWZQ@mail.gmail.com>
To: Thi Nguyen-Huu <thi.nh@winmagic.com>
Content-Type: multipart/related; boundary="0000000000004f6aaa065ce06a10"
X-Spamd-Bar: --
Message-ID-Hash: 7PV5RSO4OJEXPV3L4EEBXTENBU2YUIZ5
X-Message-ID-Hash: 7PV5RSO4OJEXPV3L4EEBXTENBU2YUIZ5
X-MailFrom: roniamdev@gmail.com
X-Mailman-Rule-Misses: dmarc-mitigation; no-senders; approved; loop; banned-address; header-match-oauth.ietf.org-0; emergency; member-moderation; nonmember-moderation; administrivia; implicit-dest; max-recipients; max-size; news-moderation; no-subject; digests; suspicious-header
CC: "刘大鹏(鹏成)" <max.ldp@alibaba-inc.com>, Yaron ZEHAVI <yaron.zehavi@rbinternational.com>, Karl McGuinness <me@karlmcguinness.com>, "oauth@ietf.org" <oauth@ietf.org>, Iman Schrock <team@emiliaprotocol.ai>
X-Mailman-Version: 3.3.10
Precedence: list
Subject: [OAUTH-WG] Re: 回复:Re: Call for interest: fine-grained authorization in OAuth (IETF 126 follow-up)
List-Id: OAUTH WG <oauth.ietf.org>
Archived-At: <https://mailarchive.ietf.org/arch/msg/oauth/MHdJDMsfViuqVbNfa8kipktcHTE>
List-Archive: <https://mailarchive.ietf.org/arch/browse/oauth>
List-Help: <mailto:oauth-request@ietf.org?subject=help>
List-Owner: <mailto:oauth-owner@ietf.org>
List-Post: <mailto:oauth@ietf.org>
List-Subscribe: <mailto:oauth-join@ietf.org>
List-Unsubscribe: <mailto:oauth-leave@ietf.org>
Hi all,
Here's the use case written up properly. It's long, and it's plain text so
it can be lifted into an issue or a draft as is.
Dapeng, I took your point. It's written as a problem statement first, not
as a pitch for a draft. I've used the mission draft as the reference design
because it's the most complete description on the list of where a goal
lives, but the questions would apply to any design with that shape, yours
included.
Thi, thanks for picking up the reserve-and-charge example. It's the one
that runs through the whole document.
Karl, the runtime enforcement companion is the piece I was missing when I
wrote the earlier version, and the document now says what it does and
doesn't reach. A permit on the charge is evaluated in the payments
vocabulary, which has an amount, an account and an order line, and no item.
So the permit is correctly granted, and the fact that it was the wrong item
lives in the join between the inventory record and the payments record,
which no per-action check can see.
That join is what I'm asking the draft to cover. The rest is smaller, a
remediation that can name a Mission, and what a gateway that sees only part
of the traffic is in the model.
I'll open issues on the repo for the specific points so they don't get lost
in the thread.
The write up follows.
Ron
Use case for Mission-Bound Authorization. A goal that spans resource
servers with unrelated vocabularies, and what happens after issuance
Ron Bar-Tor, roniam.dev, October 2026
Input to draft-mcguinness-oauth-mission-00, from the OAuth WG thread
on fine-grained authorization for AI agents
0. Why this is written against the Mission draft
My first mail asked where a goal lives when its tasks land on several
resource servers that share no authorization vocabulary. The Mission
draft answers that. The goal lives at the Authorization Server as a
Mission, the AS derives an Authority Set from the Mission Intent, a
human approves that set, and every token issued under the Mission is
checked against it at issuance by containment (sections 5, 5.1, 6, 8,
9.1). This write-up takes that answer as given and asks what the
mission layer still needs for the scenario below, which is the one I
raised and the one Iman Schrock sharpened on the list. Dapeng Liu's
point on the thread stands above all of it. The call is about whether
the problem is real and worth the group's time, not which draft
solves it, so this is written as a problem statement first. It uses
the Mission draft as its reference design because that is the most
complete description on the list of where a goal lives, and the
questions below would apply to any design that takes the same shape,
including draft-liu-agent-operation-authorization.
1. Scenario
An agent is given a goal by a user, "buy product A". This one example
runs through the whole document. The agent submits a Mission Intent
for it (section 4). The goal fans out into tasks against several
resource servers, which may sit behind a gateway. An inventory service
that reserves the item, a payments service that charges for it, a
fulfillment service that ships it, a notification service that tells
the user. Each of those resource servers owns its own authorization
semantics. Each has its own notion of what a task is, its own
constraining attributes (account, amount, item, quantity, tenant,
record class), and its own way of scoping a request, whether by scope
strings, by RAR authorization_details types of its own definition, or
by claims it expects to find in the token. The two that matter most
below are the inventory service (RS_1) and the payments service (RS_2).
Those vocabularies are not shared and, in the ordinary case, cannot
be. Services are built at different times by different teams; some
arrive through acquisition or from another business unit with an
authorization model that reflects their own history. Harmonizing the
vocabularies across the estate is a multi-year program where it is
attempted at all, and for some consumers of a given service it is not
even desirable, because the service's model is the right one for its
domain. Any design that assumes a single schema for goals and tasks
describes an estate that does not exist.
2. Actors and trust domains, in the draft's terms
Subject the user whose authority the agent acts under; also
the Approver when the user approves their own Mission.
Agent the OAuth client. Holds the plan, submits the Mission
Intent, derives tokens, issues requests. Possibly
several agents, with delegation per section 11.
Gateway an intermediary through which the agent's requests
reach the resource servers it fronts. It sees those
requests and their answers, and nothing else. Not a
call one RS makes to another on its own behalf, not an
RS outside its domain, and not the agent's traffic to
the AS unless the deployment routes it there. Nothing
in the protocol makes it see every hop; only network
policy does, and only for hops that start at the
agent. It understands no RS's vocabulary in full. The
Mission draft has no such party.
Mission Issuer
the AS that holds the Mission (section 3.1, exactly
one). In the multi-domain case a Resource AS in another
domain honors the Mission through the cross-domain
companion and never alters it.
RS_1 .. RS_n resource servers, each with its own authorization
vocabulary and its own evaluation of what it is asked.
Each enforces from the token, statelessly (8.2).
The trust domains are mutually opaque. RS_1 does not know RS_2's
schema, does not want to, and must not be required to.
3. Terms, in plain words
The IETF 126 material uses five terms and the Mission draft adds its
own. With the purchase as the example:
Goal What the user wants done, stated once at the top, "buy
product A". In the draft this is the goal member of the
Mission Intent, and once approved it is the Mission.
Task One piece of the goal, small enough to be authorized
for one RS, "reserve one unit of A at inventory" or
"charge the price of A at payments". In the draft a
task is authorized by the AS, as a derived entry of the
Authority Set, and enforced by the RS from the token;
the RS is never asked for authorization directly.
Goal state The record of how far the goal has got, which tasks
were issued, which succeeded, which were refused and
what the refusal asked for. The draft's Mission state
is active, revoked or expired (section 9), which is
lifecycle, not progress.
Discoverable schemas
How a client learns, without out-of-band agreement,
what a given RS accepts. The remediation draft does
this by refusal rather than lookup. The Mission draft
leaves it to registration and deployment policy at the
AS (section 5).
Task-in-goal evaluation
The check that a task does not exceed the goal. The
Mission draft does this at the AS, at issuance, as
containment, meaning a derived entry must be a subset
of the approved Authority Set (5.1, 8).
Mission The durable, approved, integrity-bound record at the AS
(section 7). Mission Intent is the agent's proposal;
Authority Set is what the AS derived and the Approver
consented to; the mission claim is the identifier every
derived token carries (8.1).
4. What the Mission draft settles, for this scenario
Where the goal lives. At the Mission Issuer. The per-task check
before issuance is the subset rule (5.1). My earlier write-up asked
who holds the goal and proposed the agent or the gateway; the draft
answers with the AS, and that answer works for this scenario as far
as issuance.
What an RS sees. Never the goal. Its own entries, in a single-
audience token (section 8), audience-filtered introspection (10.1),
and the mission claim as an opaque anchor it may log (8.2, 15.1).
That is a good answer to "what does the goal look like to a service
that only knows its own schema". It looks like nothing, by design.
The identifier. The mission claim, a JWT claim, not a RAR type
(1.1, 8.1); a reference for joining evidence, never a credential
(6.1). Where it rides across hops is settled in-domain and deferred
for Transaction Tokens (1.6).
Decomposition. The agent's plan is out of scope (1.3), but the agent
can propose the parts through proposed_authority in the Intent, one
entry per RS in that RS's type where the AS supports it (4, 5.3), with
the AS deriving under the subset rule. That is the shape I had
proposed, already in the draft.
5. What the scenario still asks, as three questions
Q1 After issuance, who establishes that the tasks performed under a
Mission jointly did what the Mission approved? The draft stops at
issuance and says so (1.3: revocation "does not undo actions
already completed"; 1.7: it "does not evaluate whether a given
runtime action serves the Mission's purpose"; section 4:
success_criteria "carry no machine semantics"). The runtime
enforcement companion Karl described on the list evaluates each
consequential action, with its concrete parameters, against the
Mission before it executes, which is the right check for one
action and does not reach this case, for a reason section 9
spells out. The case there is one where every action is right,
every per-action permit is correctly granted, and the Mission is
not satisfied, which neither layer sees.
Q2 When an RS refuses a task and signals remediation in its own
vocabulary (draft-ietf-oauth-rar-metadata-remediation), how does
that remediation reach the Mission? The draft forbids widening
without a fresh approval (5.1, 1.7), which is right, and has no
path from an RS's insufficient_authorization answer to a new
derivation, an Expansion successor, or a refusal that cites the
Mission.
Q3 What is the standing of a party that sees only part of the
traffic? A gateway sees the agent's calls to the RSs it fronts
and nothing an RS does downstream. It cannot be the goal holder
for the whole chain, and the draft gives it no role at all. A
deployment that has one needs to know whether it is a client,
a delegate (section 11), or nothing.
6. Requirements this scenario produces, restated
R1 A Mission's Authority Set MUST be able to carry one entry per
RS in that RS's own vocabulary, including an RS-native RAR type
the AS does not understand and a claims-shaped vocabulary, and
issue that entry to its RS. The draft allows AS-supported types
beside mission_resource_access (5.3) but cannot narrow or
project a type whose semantics it does not know, and does not
address claims-shaped RSs; scope is a projection only (8).
R2 The party that proposes the parts (the agent, or a gateway
acting for it) MUST be able to learn each RS's vocabulary
without out-of-band agreement, ahead of time from metadata or
incrementally from the RS's own remediation. The draft leaves
this to registration and deployment policy at the AS (5).
R3 The check that a task is within the goal before issuance is the
subset rule at the AS (5.1, 8). Covered. An RS MUST NOT be
required to hold the whole Authority Set to enforce its slice
(8.2, 15.1). Covered.
R4 The Mission identity MUST be carried in every request (the
mission claim, 8.1) and across hops (companion, 1.6), and MUST
be a reference, never a credential (6.1). Covered in-domain; the
chain of tasks, as opposed to the chain of actors (11), and
Transaction Token carriage are future work.
R5 A remediation from an RS MUST be expressible in that RS's
vocabulary, attachable to the Mission entry for that RS, and
judged against the Mission before anything is re-requested,
with the draft's rule that widening needs a fresh approval
kept intact. Not covered; see Q2.
R6 The authorization decision for an operation stays at the RS
(8.2). Derivation at the AS is policy about what authority may
be issued (section 5, by design); the Authority Set is a ceiling
the RS honors, and the RS's evaluation of the concrete call is
the RS's own. Nothing in the design SHOULD require the AS to
understand every RS's vocabulary beyond what it needs to derive
and narrow. Mostly covered; the gap is R1's unknown types.
R7 Local authorization of every task is NOT sufficient for Mission
satisfaction. Two tasks can each be inside the Authority Set,
carry the same mission claim, and still not jointly do what the
Approver approved. The party that asserts satisfaction MUST
have authenticated, typed information from each RS about what
it was asked and what it did, in a form that can be compared
across RSs without the RSs sharing a vocabulary. The mission
claim alone cannot carry that correspondence, and the draft's
success_criteria, rightly, carries no machine meaning. Not
covered, and the main thing this write-up asks for.
7. How existing work covers this
draft-mcguinness-oauth-mission-00
Covers the goal object, its approval, the per-task containment
check at issuance, the identifier on every token, multi-audience
Authority Sets with per-RS tokens, and delegation. Does not
cover anything after issuance (1.3, 1.7), remediation, a
gateway, or discovery. Companions cover cross-domain projection,
expansion (a successor Mission for wider authority), status
(suspend, resume, complete) and signals. The status companion is
the natural home for Q1 if it carries evidence and not only
state.
RFC 9396 (RAR)
The container for Authority Set entries. The draft's own type is
mission_resource_access; other types are allowed when an
audience consumes them (5.3).
draft-ietf-oauth-rar-metadata-remediation-00
The RS's actionable answer when a token is insufficient, in RAR
terms, with no cross-RS portability (7.3.4). The natural wire
form for Q2, if a remediation can cite the Mission it falls
outside of.
draft-ietf-oauth-identity-assertion-authz-grant-04 (ID-JAG)
Carries authorization_details across a trust domain at grant
time; the Mission cross-domain companion already depends on it.
A token exchange that returns a RAR remediation (Jeff Lombardo's
suggestion on the list) would give a broker a typed remediation
at exchange time.
draft-schrock-action-evidence-boundary-07 (AEB)
The boundary at each RS's last control point, where the exact
action is bound from material fields, the one-time credential
consumed atomically, and the outcome classified and reconciled
with authenticated evidence. Its premise is the draft's too,
that approval of a Mission is not approval of the exact action.
AEB produces exactly the typed, authenticated outcome evidence
R7 needs; it does not compare across RSs, and does not claim to.
Mission-Bound Runtime Enforcement (companion, described on the
list 1 Oct) and its AuthZEN profile
A consequential action executes only after a permit that
evaluates the action and its concrete parameters against the
approved Mission; every phase of a compound action gets its own
decision. This is per action, in the vocabulary of the RS the
action is on. It is the right second layer, and it is what makes
the case in section 9 precise. The gap is not a missing
per-action check but the join across two vocabularies that no
per-action check can perform.
draft-liu-agent-operation-authorization-02 (expired, Mar 2026)
A structured proposal of operations converted to a JWT, then a
token representing confirmed authorization for a specific
operation, enforceable at runtime. Same shape as the Mission
draft at the level of one operation; the same question arises
across operations.
Authentication of who is acting (PSEA, sender-constrained keys,
WIMSE, as raised on the list)
Whether the party performing the action is still the party
approved for it is a separate question from whether the actions
jointly satisfied the goal, and this write-up does not touch it.
Both can fail independently; the case in section 9 fails with a
fully authenticated, fully present caller.
draft-ietf-oauth-transaction-tokens-11 and RFC 8693
The cross-hop context. The Mission draft defers carrying the
mission claim there (1.6).
RFC 9470 (step-up)
The negotiation pattern's ancestor, one dimension.
8. A shape, for discussion, built on the draft
Keep everything the draft does. Add one thing on the way out. Each RS
that acts under a Mission answers with a typed, authenticated record
of what it was asked and what it did, naming the mission claim it
acted under and the material fields of the action (item, quantity,
amount, order line), the way AEB already produces at the boundary.
The draft already asks an RS to log the mission id and token jti
(8.2); this makes that record something the RS returns, or something
a status companion can collect. Whoever asserts that the Mission was
satisfied, the agent for its own plan, a Mission Status service for
the Issuer, or a gateway for the hops it can see, compares those
records across the Mission's entries against the approved Authority
Set and the Intent's goal. success_criteria stays human-readable and
unenforced, as the draft has it; satisfaction is a separate, machine-
checkable comparison of evidence, not an interpretation of prose.
The per-entry check before issuance and the cross-entry check after
are two different checks, and the second is the one a mission claim
cannot do on its own.
For Q2, the smallest change is a remediation that names the Mission,
an RS's insufficient_authorization answer that says which Mission
entry the request fell outside of, so the agent can take it to the
Issuer as a request for a successor under the expansion companion
rather than as a bare RAR request the Issuer must refuse under 5.1.
For Q3, say what a gateway is. If it is a client, it holds a Mission
of its own for the hops it makes. If it is a delegate, it holds a
narrowed token under section 11 and no Mission. Either way the draft
should say which, because deployments with a gateway will pick one.
9. A positive and a negative case
The purchase from section 1, where the Mission is "buy product A". The
Authority Set has an inventory entry (reserve one unit of A) and a
payments entry (charge the price of A), both carrying the mission
claim and a typed reference to the same order line.
Positive. RS_1 reserves A and answers with an authenticated record
naming the mission claim, the order line and the item reserved.
RS_2 charges for A and answers naming the mission claim, the order
line and the item charged. The party asserting satisfaction
compares the two records on the typed reference, finds them
consistent with the Mission, and asserts it. Nothing in this
required RS_1 to read RS_2's vocabulary or the reverse.
Negative. RS_1 reserves A. RS_2, through a stale cart, a bug, or an
adversarial agent, charges for product B. Each operation is inside
the Authority Set, because the payments entry permits a charge on
this order line and the amount matches. Both tokens issued. Both
calls succeeded. Every party that checks one operation at a time
says yes, so none of them can see that the two operations do not
add up to the purchase. The mission layer cannot see it because it
stopped at issuance. The runtime layer cannot see it either, and
this is the point to be precise about, because a per-action permit
evaluates the charge's concrete parameters against the payments
entry in the payments vocabulary, and that vocabulary has an
amount, an account and an order line and no concept of an item. The
permit is correctly granted. "Which item was this charge for" is a
fact the inventory record holds and the payments record cannot,
and the failure lives in the join between the two, which no check
on either side can perform. Only the comparison across the two
records can, and only because the records are typed, authenticated,
and name the item where their vocabulary has one; the mission claim
on both requests would have matched.
Stated in AEB's terms. AEB's MATCH state is every accepted leg of
one action denoting the same exact material action. The purchase is
two different exact actions, each bound at its own boundary, each
with its own outcome. In the negative case both are SATISFIED,
AUTHORIZED and COMMITTED, and both are OBSERVED_AS_REQUESTED,
because each was requested wrongly and performed exactly as
requested. The failure is above MATCH and above the Mission.
What this case does not cover, and the design should say so.
Correspondence is not atomicity. The reservation can be valid when
RS_1 answers and gone by the time RS_2 charges, whether expired,
released by a timeout, consumed by another order. Keeping the legs
consistent is a transaction problem, a reservation with an expiry, a
charge that names the reservation it rests on, a confirming step, a
compensation when a leg fails, and it belongs to the services and
their orchestrator. Two things follow. The records R7 requires have to
carry enough state for staleness to be detectable, not only mismatch.
And the draft should say that a mission claim is not a transaction
identifier. It joins requests to an approval for issuance and audit,
which 6.1 and 7.1 already make clear, and it promises nothing about
the ordering, timing or atomicity of the effects.
10. Open questions for the Mission draft and its companions
1. Where an RS's record of what it did goes, whether returned to
the agent in the response, exposed through the Issuer's
introspection, collected by the status companion, or all three.
R7 needs it to exist; it does not need it to be in one place.
2. Whether a gateway is a client with its own Mission, a delegate
under section 11, or outside the model. It sees every hop that
starts at the agent and none that starts at an RS.
3. Whether an RS's remediation may cite a Mission entry, so that a
refusal can become a request for a successor rather than a
request the Issuer must refuse.
4. Whether two parties holding overlapping Missions for one goal
(two agents, or agent and gateway) is a case to support or to
forbid.
5. Whether a Mission can be marked satisfied by evidence, as
distinct from completed by the Approver, and by whom.
I can turn any of this into text for the draft or its issue tracker.
On Fri, 2 Oct 2026 at 04:01, Thi Nguyen-Huu <thi.nh@winmagic.com> wrote:
> Jeff,
>
>
>
> Thanks. Yes, I've read AIMS. I replied to the adoption call in August and
> offered text for §7/§8.
>
>
>
> "Nobody's asking" was too strong — I cited Mohamad's PSEA in the same
> mail. I meant I hadn't seen it raised in this thread.
>
>
>
> Let me put the point more simply. Take a client and a server. The goal is
> for the server to stop giving data as soon as the client is no longer
> authenticated. The server decides to give data several times a second; it
> can't re-authenticate the client each time. It can check a key or a token,
> but it can't see whether the user is still at the device, or whether the
> device is still in a good state.
>
>
>
> The endpoint can. So let the endpoint decide: it checks the user and its
> own state, and uses the key - or more generally, accesses the server - only
> while all the conditions hold. It's not about the mathematics, just an
> ordinary decision the endpoint makes.
>
>
>
> So authentication - identity now - is done by the endpoint. The server
> benefits from it without any signal — yes, magic, in a way :-) — and can be
> notified if it wants. SSF is the right channel for that, and we've proposed
> presence events there.
>
>
>
> Why trust the endpoint's decision? The same way you trust any key: it's
> registered once and ideally kept in hardware.
>
>
>
> Cheers
>
>
>
> *Thi Nguyen-Huu*
>
> *CEO*
>
> winmagic.com
>
> 11-80 Galaxy Blvd.
> <https://www.google.com/maps/search/80+Galaxy+Blvd.++%7C+Toronto,+ON++%7C+M9W+4Y8++%7C+Canada?entry=gmail&source=g>
> |
> <https://www.google.com/maps/search/80+Galaxy+Blvd.++%7C+Toronto,+ON++%7C+M9W+4Y8++%7C+Canada?entry=gmail&source=g>
> Toronto, ON
> <https://www.google.com/maps/search/80+Galaxy+Blvd.++%7C+Toronto,+ON++%7C+M9W+4Y8++%7C+Canada?entry=gmail&source=g>
> |
> <https://www.google.com/maps/search/80+Galaxy+Blvd.++%7C+Toronto,+ON++%7C+M9W+4Y8++%7C+Canada?entry=gmail&source=g>
> M9W 4Y8
> <https://www.google.com/maps/search/80+Galaxy+Blvd.++%7C+Toronto,+ON++%7C+M9W+4Y8++%7C+Canada?entry=gmail&source=g>
> |
> <https://www.google.com/maps/search/80+Galaxy+Blvd.++%7C+Toronto,+ON++%7C+M9W+4Y8++%7C+Canada?entry=gmail&source=g>
> Canada
> <https://www.google.com/maps/search/80+Galaxy+Blvd.++%7C+Toronto,+ON++%7C+M9W+4Y8++%7C+Canada?entry=gmail&source=g>
>
>
>
> *From:* Lombardo, Jeff <jeffsec@amazon.com>
> *Sent:* Thursday, October 1, 2026 6:41 PM
> *To:* Thi Nguyen-Huu <thi.nh@winmagic.com>; 刘大鹏(鹏成) <max.ldp=
> 40alibaba-inc.com@dmarc.ietf.org>; Yaron ZEHAVI <
> yaron.zehavi@rbinternational.com>; Ron Bartor <roniamdev@gmail.com>; Karl
> McGuinness <me@karlmcguinness.com>
> *Cc:* oauth@ietf.org; Iman Schrock <team@emiliaprotocol.ai>; Lombardo,
> Jeff <jeffsec@amazon.com>
> *Subject:* RE: RE: [OAUTH-WG] 回复:Re: Call for interest: fine-grained
> authorization in OAuth (IETF 126 follow-up)
>
>
>
> CAUTION:This email originated from outside of the organization. Do not
> click links, open attachments or respond unless you recognize the sender
> and know that the content is safe.
>
>
>
> Hi,
>
> What do you mean by “there's a second question nobody's asking yet: is
> the one doing this action right now still the one who was approved for it?”?
>
> All the OAuth flows support Client authentication along multiple grades of
> Assurance (from Client secret to private_key JWT, going through mTLS
> authentication and JWT Assertion Grant as well as SPIFFE Client
> Authentication as part of the WIMSE work) and the binding of the issued
> token to cryptographic keys through a Proof of Possession mechanism (DPoP
> as of now, with exploration in HTTPSig) to ensure Bearer usage is coming
> for the client it was issued to originally.
>
>
>
> Any interaction with the AS and the RS can act as a stop gap in the
> acquisition of new token and the usage of new tokens.
>
> It is true that there is some signaling to do to ensure that changes in
> posture can impact, if one party consider it is a risk, the rest of the
> experience. Saying that *If those conditions stop being true, the key
> simply stops working* is too simple and magical. A key pair is always
> working from a cryptographic/mathematical standpoint, it needs to be the
> trust established on it that needs to stop working. This is what we created
> through limitations of time of any JWT (used for authentication or
> authorization), CRLs, OAuth Status Lists, and the OpenID Foundation Shared
> Signal Framework [ https://openid.net/wg/sharedsignals ] which has many
> event profiles including now a Workload Identity [
> https://openid.github.io/ssf-wise-profile/ ].
>
>
>
> Have you read about the WIMSE AIMS Draft per chance?
> https://datatracker.ietf.org/doc/draft-ietf-wimse-aims/
>
>
>
> Jeff
>
> *Jean-François “Jeff” Lombardo* | Amazon Web Services
>
>
> Principal Standards Specialist, Identity Services
> Montréal, Canada
>
>
>
> *From:* Thi Nguyen-Huu <thi.nh@winmagic.com>
> *Sent:* October 1, 2026 5:46 PM
> *To:* 刘大鹏(鹏成) <max.ldp=40alibaba-inc.com@dmarc.ietf.org>; Lombardo, Jeff <
> jeffsec@amazon.com>; Yaron ZEHAVI <yaron.zehavi@rbinternational.com>; Ron
> Bartor <roniamdev@gmail.com>; Karl McGuinness <me@karlmcguinness.com>
> *Cc:* oauth@ietf.org; Iman Schrock <team@emiliaprotocol.ai>
> *Subject:* [EXT] RE: [OAUTH-WG] 回复:Re: Call for interest: fine-grained
> authorization in OAuth (IETF 126 follow-up)
>
>
>
> *CAUTION*: This email originated from outside of the organization. Do not
> click links or open attachments unless you can confirm the sender and know
> the content is safe.
>
>
>
> *AVERTISSEMENT*: Ce courrier électronique provient d’un expéditeur
> externe. Ne cliquez sur aucun lien et n’ouvrez aucune pièce jointe si vous
> ne pouvez pas confirmer l’identité de l’expéditeur et si vous n’êtes pas
> certain que le contenu ne présente aucun risque.
>
>
>
> Hi all.
>
>
>
> If I understand correctly:
>
>
>
> Ron's example shows the problem clearly: the reservation is for product A,
> but the charge is for product B. Nothing catches this, because the check
> happens when the mission is approved, not when each action actually happens.
>
>
>
> Everyone in this thread, including Karl's design, is working on
> authorization: was this the right thing to approve, did we understand the
> request correctly, does each step still match what was approved. Karl's own
> outline even has a step called "runtime enforcement," but it's left empty.
>
>
>
> If that's a fair read, there's a second question nobody's asking yet: is
> the one doing this action right now still the one who was approved for it?
> That's authentication, not authorization. Left alone, I'd guess "runtime
> enforcement" becomes another authorization check — does this still match
> the approved scope — not a check on who is actually there.
>
>
>
> Mohamad Yossif's PSEA draft already asks for this directly: proof that a
> real human is present, and that the device is trustworthy, at the moment of
> the action — not just proof the action was authorized. Maybe the design
> team should treat this as its own topic, separate from goal-tracking and
> schema discovery.
>
>
>
> One more thought, building on Mohamad's idea: proving someone is present
> doesn't have to mean asking them to approve something again and again. A
> mission record is a decision made once, then trusted later. Checking a
> request against it only tells you the request matches what was approved —
> it says nothing about what's happened since. A different approach: give the
> agent a key that only works while the original conditions are still true —
> right person, right device. If those conditions stop being true, the key
> simply stops working. No revoke message needed, because there's nothing
> left to revoke.
>
>
>
> This isn't a competing proposal to Karl's draft, or to Mohamad's. It fills
> in the empty "runtime enforcement" step Karl already left open, and it's
> one way to get what Mohamad is asking for, without needing to ask the human
> again every time.
>
>
>
> Cheers
>
>
>
> *Thi Nguyen-Huu*
>
> *CEO*
>
> winmagic.com
>
> 11-80 Galaxy Blvd.
> <https://www.google.com/maps/search/80+Galaxy+Blvd.++%7C+Toronto,+ON++%7C+M9W+4Y8++%7C+Canada?entry=gmail&source=g>
> |
> <https://www.google.com/maps/search/80+Galaxy+Blvd.++%7C+Toronto,+ON++%7C+M9W+4Y8++%7C+Canada?entry=gmail&source=g>
> Toronto, ON
> <https://www.google.com/maps/search/80+Galaxy+Blvd.++%7C+Toronto,+ON++%7C+M9W+4Y8++%7C+Canada?entry=gmail&source=g>
> |
> <https://www.google.com/maps/search/80+Galaxy+Blvd.++%7C+Toronto,+ON++%7C+M9W+4Y8++%7C+Canada?entry=gmail&source=g>
> M9W 4Y8
> <https://www.google.com/maps/search/80+Galaxy+Blvd.++%7C+Toronto,+ON++%7C+M9W+4Y8++%7C+Canada?entry=gmail&source=g>
> |
> <https://www.google.com/maps/search/80+Galaxy+Blvd.++%7C+Toronto,+ON++%7C+M9W+4Y8++%7C+Canada?entry=gmail&source=g>
> Canada
> <https://www.google.com/maps/search/80+Galaxy+Blvd.++%7C+Toronto,+ON++%7C+M9W+4Y8++%7C+Canada?entry=gmail&source=g>
>
>
>
> *From:* 刘大鹏(鹏成) <max.ldp=40alibaba-inc.com@dmarc.ietf.org>
> *Sent:* Thursday, October 1, 2026 5:11 PM
> *To:* Lombardo, Jeff <jeffsec@amazon.com>; Yaron ZEHAVI <
> yaron.zehavi@rbinternational.com>; Ron Bartor <roniamdev@gmail.com>; Karl
> McGuinness <me@karlmcguinness.com>
> *Cc:* oauth@ietf.org; Iman Schrock <team@emiliaprotocol.ai>
> *Subject:* [OAUTH-WG] 回复:Re: Call for interest: fine-grained
> authorization in OAuth (IETF 126 follow-up)
>
>
>
> You don't often get email from max.ldp=40alibaba-inc.com@dmarc.ietf.org. Learn
> why this is important <https://aka.ms/LearnAboutSenderIdentification>
>
> CAUTION:This email originated from outside of the organization. Do not
> click links, open attachments or respond unless you recognize the sender
> and know that the content is safe.
>
>
>
> Hi Jeff and all,
>
>
>
> I think whether the problem is real and whether there is enough
> interest in the WG to work on it are separate from the question
> of which draft would be the solution.
>
> That was the purpose of
> the call for interest, and I read the replies so far as
> interest in the problem itself rather than in any particular solution.
>
>
>
> The discussion in this thread reaches a similar problem from
> several independent directions. We reached it ourselves earlier:
> draft-liu-agent-operation-authorization was presented at IETF 125.
>
> It investigates how to extend OAuth to
> express bounded work proposals from agents and the mechanism
> for approving them. It shares much of the problem and solution
> space discussed in this thread.
>
>
>
> I think the important question for the next steps and for the
> upcoming IETF 127 meeting is whether the WG has enough support
> for the problem space and enough interest to work on potential solutions.
>
> We do not need to limit ourselves to a specific draft
> at this stage. If we agree on the problem, the WG or a design
> team can work on the solution together.
>
>
>
>
>
> Regards,
>
> Dapeng
>
> ------------------原始邮件 ------------------
>
> *发件人**:*Lombardo, Jeff <jeffsec=40amazon.com@dmarc.ietf.org>
>
> *发送时间**:*Thu Oct 1 23:47:20 2026
>
> *收件人**:*Yaron ZEHAVI <yaron.zehavi@rbinternational.com>, Ron Bartor <
> roniamdev@gmail.com>, Karl McGuinness <me@karlmcguinness.com>
>
> *抄送**:*oauth@ietf.org <oauth@ietf.org>, Iman Schrock <
> team@emiliaprotocol.ai>
>
> *主**题**:*[OAUTH-WG] Re: Call for interest: fine-grained authorization in
> OAuth (IETF 126 follow-up)
>
> Agreed with Yaron, I thought I provided the reference. It seems not.
>
> @Ron : The discussion is the right one, it just that I think we should
> have it as part of the Mission Draft instead of yet another Draft.
>
>
>
> Jeff
>
>
>
> *Jean-François “**Jeff**” Lombardo* |Amazon Web Services
>
>
> Principal Standards Specialist, Identity Services
> Montréal, Canada
>
>
>
> *From:* Yaron ZEHAVI <yaron.zehavi@rbinternational.com>
> *Sent:* October 1, 2026 1:43 AM
> *To:* Ron Bartor <roniamdev@gmail.com>; Karl McGuinness <
> me@karlmcguinness.com>
> *Cc:* Lombardo, Jeff <jeffsec@amazon.com>; oauth@ietf.org; Iman Schrock <
> team=40emiliaprotocol.ai@dmarc.ietf.org>
> *Subject:* RE: [EXT] [OAUTH-WG] Re: Call for interest: fine-grained
> authorization in OAuth (IETF 126 follow-up)
>
>
>
> *CAUTION*: This email originated from outside of the organization. Do not
> click links or open attachments unless you can confirm the sender and know
> the content is safe.
>
>
>
> *AVERTISSEMENT*: Ce courrier électronique provient d’un expéditeur
> externe. Ne cliquez sur aucun lien et n’ouvrez aucune pièce jointe si vous
> ne pouvez pas confirmer l’identité de l’expéditeur et si vous n’êtes pas
> certain que le contenu ne présente aucun risque.
>
>
>
> Dear Ron,
>
>
>
> I think the goal enforcement you’re seeking may be found in@Karl
> McGuinness <me@karlmcguinness.com>’s draft:Mission-Bound Authorization
> for OAuth 2.0
> <https://datatracker.ietf.org/doc/html/draft-mcguinness-oauth-mission-00>.
>
> It adds an AS-persisted durable higher level mission intent object,
> against which authority requests are evaluated at token issuance time.
>
>
>
> It should empower the authorization server to evaluate if reservation A
> and payment B are within the scope of the approved mission, apply policies
> to validate they’re related and if so issue the tokens for the relevant
> audience resource servers.
>
>
>
> Regards,
>
> Yaron
>
>
>
>
>
> Classification: GENERAL
>
> *From:* Ron Bartor <roniamdev@gmail.com>
> *Sent:* Thursday, October 1, 2026 3:58 AM
> *To:* Yaron ZEHAVI <yaron.zehavi@rbinternational.com>
> *Cc:* Lombardo, Jeff <jeffsec@amazon.com>;oauth@ietf.org; Iman Schrock <
> team=40emiliaprotocol.ai@dmarc.ietf.org>
> *Subject:* Re: [OAUTH-WG] Re: Call for interest: fine-grained
> authorization in OAuth (IETF 126 follow-up)
>
>
>
> You don't often get email fromroniamdev@gmail.com. Learn why this is
> important <https://aka.ms/LearnAboutSenderIdentification>
>
> This message is from an external sender - be cautious, particularly with
> links and attachments.
>
>
>
> Hi Jeff, Yaron,
>
> Yaron, thanks for the correction, and good to meet you. I was citing the
> old draft, my mistake. I'll cite draft-ietf-oauth-rar-metadata-remediation
> from here on and stop saying "per-RS discovery", since that's not what it
> does. The way I read it now, the metadata endpoint tells a client what the
> AS can issue, and what a particular resource actually needs you find out by
> calling it and getting the remediation back. I also take the gateway point.
> It relays the call and relays the remediation and doesn't need to know
> anything about the downstream types to do that.
>
> Jeff, agreed on the rest. Each RS says what it needs in RAR terms, the AS
> normalizes, and if there's a broker in front it adds one more layer and
> normalizes for its own client. That's the per-RS problem, and I think the
> draft handles it well. My write-up says as much.
>
> What I'm trying to point at is one step above that, and a concrete case is
> clearer than my first mail was. Say the goal is "buy product A". It turns
> into a reservation at an inventory service and a charge at a payments
> service (two RSs with their own vocabularies, and quite possibly two
> different ASs). Each RS remediates in its own terms, each remediation gets
> normalized correctly, each request gets authorized. Except the reservation
> is for A and the charge is for B. Nothing in that path compares the two,
> and I don't think anything in it can, because comparing them needs the
> goal, and the AS that has authority over the payments service doesn't hold
> the goal. It holds a grant for one service, not what the user was trying to
> do across several.
>
> Yaron's correction actually makes this a bit sharper. If there's no
> discovery ahead of time, whoever holds the goal can't decompose it into
> typed parts up front. It only learns what each RS needs when that RS
> refuses. So the question "is this remediation still within the goal" has to
> be asked on every remediation as it comes back, and then once more across
> all of them at the end. The gateway relays those without reading them,
> which is the right thing for a gateway. But somebody still has to read them
> against the goal, and that's the party I'd like the design team to name,
> whether it ends up being the client, the AS, the broker, or each RS for its
> own slice. If the answer is "the broker, when there is one", fine, that's
> an answer, but I'd rather it was said out loud, because it means the goal,
> and whatever evidence is needed to judge against it, lives at the
> intermediary.
>
> Jeff, I think your token exchange idea fits here. If an exchange can hand
> back a RAR remediation, the broker gets a typed remediation at token time,
> in the RS's own vocabulary, without having to be refused by the RS first.
> For the per-RS half of this across a broker hop that's probably the
> cleanest carrier, and I'll put it in the write-up next to ID-JAG.
>
> Ron
>
>
>
> On Wed, 30 Sept 2026 at 08:18, Yaron ZEHAVI <
> yaron.zehavi@rbinternational.com> wrote:
>
> Hello everyone,
>
>
>
> @Lombardo, Jeff <jeffsec@amazon.com> thanks for the shout out.
>
> @Ron Bartor <roniamdev@gmail.com>, pleased to meet you
>
>
>
> One comment, OAuth 2.0 RAR Metadata and Error Remediation does not offer
> per-RS metadata discovery.
>
> This idea was explored in earlier drafts:
>
>
> https://www.ietf.org/archive/id/draft-zehavi-oauth-rar-metadata-03.html#name-oauth-20-protected-resource
>
>
>
> And removed following feedback from WG members.
>
> The idea was add to RFC 9728 RS metadata:
>
>
>
> "authorization_details_types_required": {
>
> "oneOf": ["payment_initiation", "payment_approval",
>
> "beneficiary_designation"]
>
> }
>
>
>
> With Boolean expressions allOf, and, or.
>
> This was deemed not very helpful as:
>
> - Different resources have different requirements (but in MCP the
> resource is not expressed on the http path)
> - Lots of logic which in the end doesn’t tell the client exactly which
> RAR types to request.
>
>
>
> Then the next idea was include the Boolean logic in the WWW-Authenticate
> header. Opted instead to return the actionable RAR object to remediate the
> failure.
>
>
>
> That being said, I think the task of RS-Gateways was made simpler.
>
> They don’t need to have foreknowledge to combine downstream RS metadata.
>
>
>
> Instead, client makes a request, a failing downstream RS may return the
> remediation payload. Gateway relays the calls without requiring additional
> metadata.
>
>
>
> Yaron
>
>
>
>
>
> Classification: GENERAL
>
> *From:* Lombardo, Jeff <jeffsec@amazon.com>
> *Sent:* Wednesday, September 30, 2026 4:08 PM
> *To:* Ron Bartor <roniamdev@gmail.com>
> *Cc:* oauth@ietf.org; Yaron ZEHAVI <yaron.zehavi@rbinternational.com>;
> Iman Schrock <team=40emiliaprotocol.ai@dmarc.ietf.org>; Lombardo, Jeff <
> jeffsec@amazon.com>
> *Subject:* RE: [OAUTH-WG] Re: Call for interest: fine-grained
> authorization in OAuth (IETF 126 follow-up)
>
>
>
> This message is from an external sender - be cautious, particularly with
> links and attachments.
>
>
>
> Hi Ron,
>
> Speaking under the supervision of Yaron. I am not sure I follow the multi-RS
> decomposition problem you try to point at here.
>
>
>
> draft-zehavi-oauth-rar-metadata now
> draft-ietf-oauth-rar-metadata-remediation [
> https://datatracker.ietf.org/doc/draft-ietf-oauth-rar-metadata-remediation/]
> allows each RS to express the RAR-based mediated response along what the AS
> that has authority can speak in terms of RAR schema.
>
>
>
> It means that it is up to each AS to normalize the mediated response, If
> brokered through an API Gateway, this is just adding another layer and it
> is now up to the Gateway / Broker to normalize it for its client.
>
>
>
> As ID-JAG supports RAR, there could be an extension to Token Exchange to
> return a RAR remediation to ease this path.
>
>
>
> Jeff
>
>
>
>
>
> *Jean-François “**Jeff**” Lombardo* |Amazon Web Services
>
>
> Principal Standards Specialist, Identity Services
> Montréal, Canada
>
>
>
> *From:* Ron Bartor <roniamdev@gmail.com>
> *Sent:* September 29, 2026 7:09 PM
> *To:* Iman Schrock <team=40emiliaprotocol.ai@dmarc.ietf.org>
> *Cc:* oauth@ietf.org
> *Subject:* [EXT] [OAUTH-WG] Re: Call for interest: fine-grained
> authorization in OAuth (IETF 126 follow-up)
>
>
>
> *CAUTION*: This email originated from outside of the organization. Do not
> click links or open attachments unless you can confirm the sender and know
> the content is safe.
>
>
>
> *AVERTISSEMENT*: Ce courrier électronique provient d’un expéditeur
> externe. Ne cliquez sur aucun lien et n’ouvrez aucune pièce jointe si vous
> ne pouvez pas confirmer l’identité de l’expéditeur et si vous n’êtes pas
> certain que le contenu ne présente aucun risque.
>
>
>
> Hi Dapeng, all,
>
>
>
> I support continuing this direction and would like to contribute to the
> design team.
>
>
>
> I wasn't in the room at 126, so apologies if this was covered, but I'd
> like to put one scenario on the table for the "goal state management" and
> "discoverable schema" pillars. I operate an enterprise API gateway that
> fronts MCP and agent traffic. A single agent request routinely fans out to
> several downstream resource servers, and each of those RSs owns its own
> authorization semantics: its own notion of what a task is, what the
> constraining attributes are, and how a request is scoped. Those semantics
> are not shared across services, and in some cases cannot be. A concrete
> example is an RS that was integrated from a separate business unit; its
> authorization model reflects its own history and is materially different
> from the rest of our services, and harmonizing it is neither quick nor, for
> some of its consumers, desirable.
>
>
>
> In that setting, per-RS schema discovery (as in
> draft-zehavi-oauth-rar-metadata) solves the local problem but not the
> cross-RS one: the goal has to be carried across trust domains that are
> mutually opaque, and something has to decide whether each downstream task
> is still within the goal without any RS seeing the others' vocabulary. The
> slides put "task in goal" evaluation out of scope, which I understand, but
> I think the design team should at least be explicit about where that
> evaluation lives in a multi-RS decomposition (client, AS,
> gateway/intermediary, or each RS independently) and what the goal-state
> representation needs to look like so that it remains meaningful to an RS
> that only understands its own schema.
>
>
>
> I'd be glad to write this up as a concrete use case, and I think it fits
> naturally with the RS-to-client signaling work already in the WG
> (txn-challenge, client-challenge, RAR-metadata) rather than requiring new
> machinery.
>
>
>
> Best regards,
>
> Ron Bar-Tor
>
>
>
> On Tue, Sep 29, 2026 at 3:13 PM Iman Schrock <team=
> 40emiliaprotocol.ai@dmarc.ietf.org> wrote:
>
> Hi Dapeng,
>
> I support continuing this work. One boundary I would like the design team
> to test: a goal-level authorization can help a client negotiate what an RS
> requires, but it is not approval of the eventual exact action. A changed
> payment destination after negotiation, or a lost response after provider
> entry, needs its own decision and recovery rule.
>
> We have published AEB
> <https://datatracker.ietf.org/doc/draft-schrock-action-evidence-boundary/>
> and an open vector pack
> <https://github.com/emiliaprotocol/emilia-protocol/tree/main/conformance/aeb-1>
> for that boundary. I would be glad to bring a small cross-system case to
> the design team and see where it belongs alongside existing OAuth work.
>
> Best,
> Iman
>
>
>
> On Wed, 30 Sep 2026 00:19:52 +0800, "刘大鹏(鹏成)" <max.ldp=
> 40alibaba-inc.com@dmarc.ietf.org> wrote:
>
> Hello all, During IETF 126 we had a discussion on how OAuth can support
> fine-grained authorization for AI agents [1]. The topic generated a lot of
> interest in the room, and the poll showed support for continuing the
> discussion [2]. A quick summary of the problem space as presented in the
> session [3]: 1. Enabling clients to construct fine-grained authorization
> requests using: a. Goal state management, b. Discoverable domain-specific
> schemas for task and authorization definition, c. Task-in-goal evaluation
> requests 2. Establishing a negotiation pattern that allows a resource
> server to guide clients towards the required authorization The session also
> proposed forming a design team to explore this space [3]. Before IETF 127,
> we would like to hear from the WG: - Is this a direction the community is
> interested in pursuing? Replies on the list will help shape the discussion
> at IETF 127. [1]
> https://datatracker.ietf.org/meeting/126/materials/minutes-126-oauth-202607240700-00
> [2]https://datatracker.ietf.org/doc/polls-126-oauth-202607240900/ [3]
> https://datatracker.ietf.org/meeting/126/materials/slides-126-oauth-sessb-policy-language-binding-for-oauth-00
> Best regards,Dapeng Liu
>
> _______________________________________________
> OAuth mailing list -- oauth@ietf.org
> To unsubscribe send an email to oauth-leave@ietf.org
>
> This message and any attachment ("the Message") are confidential. If you
> have received the Message in error, please notify the sender immediately
> and delete the Message from your system, any use of the Message is
> forbidden. Correspondence via e-mail is primarily for information purposes.
> RBI neither makes nor accepts legally binding statements via e-mail unless
> explicitly agreed otherwise. Information pursuant to § 14 Austrian
> Companies Code: Raiffeisen Bank International AG; Registered Office: Am
> Stadtpark 9
> <https://www.google.com/maps/search/Am+Stadtpark+9?entry=gmail&source=g>,
> 1030 Vienna, Austria; Company Register Number: FN 122119m at the Commercial
> Court of Vienna (Handelsgericht Wien).
>
> This message and any attachment ("the Message") are confidential. If you
> have received the Message in error, please notify the sender immediately
> and delete the Message from your system, any use of the Message is
> forbidden. Correspondence via e-mail is primarily for information purposes.
> RBI neither makes nor accepts legally binding statements via e-mail unless
> explicitly agreed otherwise. Information pursuant to § 14 Austrian
> Companies Code: Raiffeisen Bank International AG; Registered Office: Am
> Stadtpark 9
> <https://www.google.com/maps/search/Am+Stadtpark+9?entry=gmail&source=g>,
> 1030 Vienna, Austria; Company Register Number: FN 122119m at the Commercial
> Court of Vienna (Handelsgericht Wien).
>
>
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Arun Thallapelly
- [OAUTH-WG] Call for interest: fine-grained author… 刘大鹏(鹏成)
- [OAUTH-WG] Re: Call for interest: fine-grained au… Iman Schrock
- [OAUTH-WG] Re: Call for interest: fine-grained au… Ron Bartor
- [OAUTH-WG] Re: Call for interest: fine-grained au… Iman Schrock
- [OAUTH-WG] Re: Call for interest: fine-grained au… Ron Bartor
- [OAUTH-WG] Re: Call for interest: fine-grained au… Lombardo, Jeff
- [OAUTH-WG] Re: Call for interest: fine-grained au… Lombardo, Jeff
- [OAUTH-WG] Re: Call for interest: fine-grained au… Yaron ZEHAVI
- [OAUTH-WG] Re: Call for interest: fine-grained au… Ron Bartor
- [OAUTH-WG] Re: Call for interest: fine-grained au… Yaron ZEHAVI
- [OAUTH-WG] Re: Call for interest: fine-grained au… Lombardo, Jeff
- [OAUTH-WG] Re: Call for interest: fine-grained au… Alex Babeanu
- [OAUTH-WG] Re: Call for interest: fine-grained au… Karl McGuinness
- [OAUTH-WG] 回复:Re: Call for interest: fine-grained… 刘大鹏(鹏成)
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Thi Nguyen-Huu
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Karl McGuinness
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Thi Nguyen-Huu
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Yaron ZEHAVI
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Ron Bartor
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Ron Bartor
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Kieran Sweeney
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Ron Bartor
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Karl McGuinness
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Ron Bartor
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Blake Morrison
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Karl McGuinness
- [OAUTH-WG] Re:Re: Call for interest: fine-grained… 刘大鹏(鹏成)
- [OAUTH-WG] Re:Re: Call for interest: fine-grained… Blake Morrison
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Yaron ZEHAVI
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Blake Morrison
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Yaron ZEHAVI
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Blake Morrison
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Yaron ZEHAVI
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Blake Morrison
- [OAUTH-WG] 回复:Re: 回复:Re: Call for interest: fine-… 刘大鹏(鹏成)
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Omri Gazitt
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Yaron ZEHAVI
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Omri Gazitt
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Kieran Sweeney
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Ron Bartor
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Kieran Sweeney
- [OAUTH-WG] Re: Call for interest: fine-grained au… Kieran Sweeney
- [OAUTH-WG] Re: Call for interest: fine-grained au… Ron Bartor
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Lombardo, Jeff
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Thi Nguyen-Huu
- [OAUTH-WG] Re: 回复:Re: Call for interest: fine-gra… Ron Bartor
- [OAUTH-WG] Re: [OAUTH-WG]回复:Re: Call for interest… Meiling Chen
- [OAUTH-WG] Re: Call for interest: fine-grained au… Mohamad Khalil Yossif
- [OAUTH-WG] Re: Call for interest: fine-grained au… Iman Schrock
- [OAUTH-WG] Re: Call for interest: fine-grained au… Mohamad Khalil Yossif
- [OAUTH-WG] Re: Call for interest: fine-grained au… Iman Schrock