| Internet-Draft | Mission Runtime OAuth Profile | November 2026 |
| McGuinness | Expires 21 May 2027 | [Page] |
This document profiles Mission-Bound Runtime Enforcement for OAuth 2.0 access tokens. It specifies how validated token claims and token introspection results supply the runtime decision inputs, how OAuth mechanisms provide Mission state observations, and how protected resource metadata conveys runtime classification floors. Enforcement semantics are defined by Mission-Bound Runtime Enforcement; token semantics are defined by Mission-Bound Authorization for OAuth 2.0.¶
This note is to be removed before publishing as an RFC.¶
The latest revision of this draft can be found at https://mcguinness.github.io/mission-bound-authorization/draft-mcguinness-mission-runtime-oauth.html. Status information for this document may be found at https://datatracker.ietf.org/doc/draft-mcguinness-mission-runtime-oauth/.¶
Source for this draft and an issue tracker can be found at https://github.com/mcguinness/mission-bound-authorization.¶
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.¶
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.¶
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."¶
This Internet-Draft will expire on 21 May 2027.¶
Copyright (c) 2026 IETF Trust and the persons identified as the document authors. All rights reserved.¶
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Revised BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Revised BSD License.¶
Mission-Bound Runtime Enforcement [I-D.draft-mcguinness-mission-runtime] (the "runtime core") defines a binding-neutral decision contract: an established Mission reference, an effective-authority source, an active predicate and freshness bound, a subject and actor, an action and resource with normalized parameters, local-policy intersection, an authenticated permit or deny, an execution boundary, and fail-closed behavior on anything the deployment does not understand. It names these as abstract roles, so a non-OAuth binding can implement it without importing OAuth semantics. Mission-Bound Authorization for OAuth 2.0 [I-D.draft-mcguinness-oauth-mission] (the "issuance profile") defines the Mission-bound access token: its claims, its issuance and delegation, and the validation a resource server applies to it.¶
Neither says which validated token values supply which runtime inputs, which OAuth mechanisms observe Mission state, or how a protected resource publishes its classification floor. This document answers those three questions for a deployment whose Mission-bound credential is an OAuth access token. Three specifications divide the work:¶
| Specification | Owns |
|---|---|
| Issuance profile [I-D.draft-mcguinness-oauth-mission] | Token claims, issuance, delegation, and baseline resource server validation |
| Runtime core [I-D.draft-mcguinness-mission-runtime] | Decision inputs, enforcement, freshness requirements, permits, and evidence obligations |
| This document | The mapping between them, OAuth state-source integration, and classification metadata |
This document defines the processing rules that mapping needs and one protected resource metadata member (Section 6). It cites the runtime core's invariants, failure modes, and evidence requirements and the issuance profile's token validation rather than restating them.¶
A decision API is a separate choice. The AuthZEN profile ([I-D.draft-mcguinness-mission-authzen]) carries the runtime core's decision between a PEP and a PDP whatever credential supplied its inputs, and a deployment using this document does not need it. A deployment on a different Mission substrate supplies its own credential profile and uses the runtime core unchanged ([I-D.draft-mcguinness-mission-substrate]).¶
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶
This specification uses the terms "access token", "authorization server", "client", "protected resource", "resource owner", and "Resource Server" from OAuth 2.0 [RFC6749] through the terminology incorporated by [I-D.draft-mcguinness-oauth-mission]. It uses Policy Enforcement Point (PEP), Policy Decision Point (PDP), established Mission, decision, Resource policy, consequential action, high-consequence class, Enforcement Scope Statement, and the action-class names as defined by the runtime core ([I-D.draft-mcguinness-mission-runtime]).¶
The values a PEP has established for a presented access token under Section 3: the claims of a validated JWT access token, or the members of the introspection response for a presented opaque token.¶
The authority the presented token itself carries: for a Mission-bound
token, the authorization_details in the validated credential
context; for an ordinary token joined to a Mission under an
externally established reference, the authority it carries as issued
(Section 5.1).¶
The Mission's approved Authority Set narrowed by whatever narrowing mechanism the deployment runs, as the runtime core's authority input defines it; the issuance profile's Status companion names it the Effective Authority Set ([I-D.draft-mcguinness-oauth-mission-status]).¶
The runtime decision is downstream of ordinary access token validation. Before using a token's Mission, authority, subject, client, actor, or confirmation-key values as decision inputs, the PEP MUST establish that the access token is valid for the protected resource and request, including its audience and, for a sender-constrained token, proof of possession of the confirmation key.¶
The issuance profile defines that validation for both of its token forms; this document adds only the audience rule of Section 3.1:¶
A JWT access token is validated under the issuance profile's resource server rules, which apply [RFC9068] and verify any sender-constraint binding ([I-D.draft-mcguinness-oauth-mission], Section "Resource Server Enforcement").¶
An opaque access token is resolved under the issuance profile's
introspected token consumption mode: introspection before each use,
an active response, the resource server's own identity in aud,
and a sender-constraint binding verified locally
([I-D.draft-mcguinness-oauth-mission], Section "Introspected Token
Consumption").¶
Each value the runtime decision uses is read from the resulting validated credential context (Section 4).¶
For an action in a high-consequence class, the runtime core requires a
sender-constrained acting credential
([I-D.draft-mcguinness-mission-runtime], Section "Credential Custody
and Mediated Execution"). The cnf claim of a validated JWT, or the
cnf member of the introspection response for an opaque token
(Section 6.2 of [RFC9449], Section 3.2 of [RFC8705]), supplies the
sender-constraint binding: a DPoP key (jkt), proved with a DPoP proof
checked per Section 7.1 of [RFC9449], or a client certificate
thumbprint (x5t#S256), proved over mutual TLS per
Section 3 of [RFC8705]. Proof over mutual TLS means the certificate
authenticated on the request's own TLS connection matches the bound
thumbprint; a certificate merely received with the request, or a
thumbprint match against a certificate the connection did not
authenticate, establishes no possession. An access token whose validated credential
context carries no such binding is a bearer token and does not meet that
requirement, and a binding whose proof the PEP has not verified for the
request supplies no confirmation key to the decision. The requirement
attaches to the action's class, not to issuance: the issuance profile's
level for the primary access token is unchanged
([I-D.draft-mcguinness-oauth-mission], Section "Mission-Bound Token
Issuance").¶
Under the issuance profile alone, introspection freshness is per use:
each response is an observation for one decision, not a cacheable state
assertion ([I-D.draft-mcguinness-oauth-mission], Section "Mission
State via Token Introspection"). A deployment that runs the Mission
Status profile can have the Mission issuer's introspection projection
carry fresh_until, until which the reported Mission state can be
relied on without re-checking
([I-D.draft-mcguinness-oauth-mission-status]). That caches Mission
state only. Token validation and the credential authority are not
cached: an opaque token is resolved by introspection before each use,
and each response remains one observation, never a cacheable authority
record ([I-D.draft-mcguinness-oauth-mission], Section "Introspected
Token Consumption").¶
A PEP MUST NOT ask a PDP to authorize an action from unverified token claims. If token validation fails, the PEP MUST refuse before runtime Mission evaluation. When the PEP is an OAuth resource server, it uses the normal OAuth error behavior for the protected resource (for example, Bearer token errors under [RFC6750]); this document defines no new OAuth error code.¶
A Mission reference reaches the decision in one of two ways. A
credential-carried reference is the mission claim of a validated JWT,
or the mission member of the introspection response for an opaque
token. An externally established reference is supplied outside the
token and verified against the acting credential under a join a
binding profile defines, as the runtime core's Mission Binding
Establishment requires ([I-D.draft-mcguinness-mission-runtime]): the
Mission Authority Server's Mission Join is one
([I-D.draft-mcguinness-mission-authority-server]), and this document
defines none of its own. If the
deployment requires Mission governance for the protected operation and
neither reference is established, the PEP MUST refuse. An external
reference never substitutes for the claim where the protected resource
requires Mission-bound tokens: a resource that advertises
mission_bound_authorization_required rejects a token that lacks the
mission claim ([I-D.draft-mcguinness-oauth-mission], Section
"Protected Resource Metadata"), whatever reference the deployment could
establish externally.¶
On this profile, the runtime core's boundary identifier
([I-D.draft-mcguinness-mission-runtime], Section "Enforcement
Boundary Identifier") is the resource identifier of the protected
resource the PEP guards (Section 1.2 of [RFC9728]), spelled exactly
as its protected resource metadata resource value
(Section 2 of [RFC9728]); the Enforcement Scope Statement declares
the boundary by that string. Trusted static configuration MAY supply
the value. A PEP that retrieves the metadata validates it under
Section 3.3 of [RFC9728] and compares identifiers as
Section 6 of [RFC9728] requires.¶
The PEP establishes the boundary before it reads aud, then checks
that aud names it; this is how the resource server determines the
identifier it expects for itself (Section 4 of [RFC9068]). aud, in
a validated JWT or an introspection response, is a string or an array
of strings (Section 4.1.3 of [RFC7519]), and the issuance profile
lets it name a resource server, API, or security domain
([I-D.draft-mcguinness-oauth-mission], Section "Mission-Bound Token
Issuance"). An aud value names the boundary when it equals the
boundary identifier or is an audience alias for it. An audience alias
is trusted deployment configuration, scoped to one validated issuer
and token profile, that maps an aud value to exactly one protected
resource identifier; several aliases MAY map to one boundary. The PEP
MUST NOT accept an alias mapping that the request or the token
supplies.¶
The PEP MUST refuse, as a token-validation failure before any PDP
decision (Section 3), a token whose aud is absent or
malformed or names the boundary neither directly nor through an
alias, and a token whose issuer and token profile have an alias that
maps to more than one protected resource. For an array, the PEP
checks that at least one member names the boundary. It MUST NOT
select a boundary from aud: not the first member, not a
serialization of the array, and not whichever member an alias maps
first. Member order and duplicate members that name the same
boundary do not change the result. A member that names another
recipient does not authorize this boundary and does not by itself
invalidate the token. A token profile whose audience rule is
stricter, for example one that requires a single audience, keeps that
rule.¶
The decision, the permit binding, and the evidence carry the boundary
identifier, never an aud value or an alias. Across a boundary
identifier change ([I-D.draft-mcguinness-mission-runtime], Section
"Changing a Boundary Identifier"), the new statement version declares
the aliases per issuer and token profile. A token whose aud names
the old identifier stays acceptable through a declared alias, under
its unchanged validation and lifetime, and the alias never rewrites an
entry's resource.¶
The runtime core states its decision inputs, permit binding, and required evidence fields as abstract roles ([I-D.draft-mcguinness-mission-runtime]). This section is their complete, normative realization on this profile. Each value is an established runtime input: a credential-derived value is read from the validated credential context (Section 3), as a JWT claim or the introspection response member of the same name, and the boundary identifier comes from the PEP's trusted routing or configuration (Section 3.1).¶
| Runtime-core role or input | OAuth realization |
|---|---|
| Established Mission reference | The mission claim's id and issuer [I-D.draft-mcguinness-oauth-mission], or an externally established reference (Section 3) |
| Authenticated subject |
sub
|
| Client identity |
client_id
|
| Immediate actor | The current actor in act when act is present; otherwise the client client_id names |
| Actor-delegation chain |
act, when delegation is in effect |
| Attester-verified actor context | Instance Context: the client_instance claim or introspection member ([I-D.draft-mcguinness-oauth-client-instance-id]), validated as that profile's Context Consumer requires; it identifies the instance that obtained the token and grants no authority, and attributing a presentation to it requires the presenter association of [I-D.draft-mcguinness-oauth-client-instance-id], Section 7.5: an instance-unique key under direct Client Attestation validation, and authenticated provenance for context preserved from an input token |
| Sender-constraint confirmation |
cnf, verified during validation (Section 3) |
| Boundary identifier | The resource identifier of the protected resource the PEP guards, which aud names directly or through an alias (Section 3.1) |
| Authority entry | The applicable entry of the credential authority for a Mission-bound token; under a join, the applicable entry of the Mission's authority (Section 5.1) |
| Current effective authority | The approved Authority Set as narrowed, from a source that reports the narrowing (Section 5.1) |
| Credential issuer |
iss
|
| Credential expiry |
exp
|
The Mission's expires_at (time input) |
The mission claim's expires_at member where present, or a Mission state source that reports the Mission's expiry |
Mission state version (mission_state_version) |
The Mission Status version member, the Mission's state version ([I-D.draft-mcguinness-oauth-mission-status]) |
Four pairs in the table are related but distinct inputs:¶
client_id and act. The issuance profile gives client_id its
ordinary meaning, the OAuth client that requested the token, and
carries delegation in act ([I-D.draft-mcguinness-oauth-mission],
Section "Resource Server Enforcement"). When delegation is in
effect, the PDP MUST evaluate the authenticated act claim as part
of the runtime actor context and refuse a chain that is missing or
malformed; when an act claim is present, the PDP MUST NOT treat
client_id alone as the immediate actor.¶
aud and the boundary identifier. Validation establishes that
aud names the boundary (Section 3.1); the decision, the
permit binding, and the evidence carry the boundary identifier. It
is not the action's target: the request's transport URL, the
Resource-policy type and id, and the operation target an entry
is matched against (Section 5.1) are distinct
values, and an alias never rewrites an entry's resource or makes
it match another target.¶
The credential authority and the current effective authority. The action falls within both (Section 5.1).¶
Credential expiry and Mission expiry. The PDP MUST refuse if the decision
context indicates the token is expired (exp). Where the validated
mission claim carries the expires_at member, the PEP or PDP MAY
refuse actions past that instant without consulting a state source:
the value is an immutable commitment and a ceiling only, carrying no
liveness, so the runtime core's only-active rule and freshness
requirements apply unchanged. Where a Mission state source separately
reports the Mission terminated with reason expired, or exposes
the Mission's expires_at, the PDP MUST refuse on it independent of
the token's own exp: the baseline mission claim need not carry
expires_at, and OAuth token introspection [RFC7662] does not
itself surface it. The issuance
profile caps the exp of every token the Mission Issuer derives at
the Mission's expires_at ([I-D.draft-mcguinness-oauth-mission],
Section "Mission-Bound Token Issuance"), so the exp check enforces
the Mission's expiry transitively, as the runtime core's time input
expects. A profile that makes the expires_at member REQUIRED for
the credentials it governs, such as the Issuance Grant profile
([I-D.draft-mcguinness-oauth-mission-issuance-grant]), lets a
validator check that cap directly.¶
These realize the runtime core's decision inputs ([I-D.draft-mcguinness-mission-runtime]); the requirements themselves, including that no runtime input expands authority beyond the issued authority, are the runtime core's.¶
The Mission reference is id and issuer. Neither authority_hash nor
intent_hash is carried on the baseline mission claim or the default
introspection projection, so a PDP has them for the runtime core's
evidence only with direct Mission-record access, introspection's
authority_hash disclosure privilege
([I-D.draft-mcguinness-oauth-mission]), or the Local Approved-Set
Verification profile
([I-D.draft-mcguinness-oauth-mission-approved-set-verification],
Section "Local Approved-Set Verification"). A deployment that needs
authority_hash as a commitment proof rather than an audit correlator
obtains it from the Mission issuer under the last of these. The runtime
core's permit binding and required decision evidence record the roles
above; their serialization is defined by the runtime core and the
decision-API profile in use, for example
[I-D.draft-mcguinness-mission-authzen].¶
The instance specification defines how to validate context and associate it with a presenter; this profile selects that association when Resource policy or the Enforcement Scope Statement requires instance attribution. The PEP validates the credential, Instance Context, and current proof under [I-D.draft-mcguinness-oauth-client-instance-id], Sections 7.3 and 7.5, before supplying the resulting identity to the PDP. The PDP relies on that PEP only under the decision API's authenticated trust boundary. A policy match on the resulting identity still grants no authority beyond the decision's other bounds.¶
Where the path requires instance attribution, the PEP MUST refuse the
request if context is absent or invalid, or if its association with the
current presenter cannot be established. This is a credential-validation
failure, using the instance specification's Section 7.6 invalid_token
response and the runtime profile's pre-decision refusal evidence. An
instance-unique key alone does not establish the association for
upstream context. Optional context that establishes only participation
can inform provenance evidence, but does not identify the current actor
instance. This requirement is conditional on the path's policy, not a
requirement that every Mission deployment use instance attribution.¶
A resource owner can carry its classification minimums to any PDP through its protected resource metadata [RFC9728], the OAuth realization of the runtime core's classification-floor rule ([I-D.draft-mcguinness-mission-runtime]):¶
mission_action_class_floors:OPTIONAL JSON object. Each member name is an action identifier from
the resource's actions vocabulary
([I-D.draft-mcguinness-oauth-mission-resource-access]); an
action-family identifier, in that profile's action-family form, sets the
floor for every action in the family. Each value is the minimum
runtime action class for the mapped action: one of
consequential_read, consequential_write, irreversible_action,
external_commitment, or privileged_administration: this
document's identifiers for the runtime core's Consequential read,
Consequential write, Irreversible action, External commitment, and
Privileged administration classes.¶
A PDP with access to the resource's metadata MUST NOT classify a mapped action below its floor. The member is the interoperable carriage of the Resource-policy minimum the runtime core's classification floor already binds; it raises, and never lowers, an action's class. A PDP that does not recognize a mapped value MUST treat it as naming a high-consequence class.¶
For the ERP resource of the runtime core's worked examples:¶
{
"resource": "https://erp.example.com",
"mission_action_class_floors": {
"journal-entries.read": "consequential_read",
"journal-entries.write": "irreversible_action"
}
}
¶
A deployment conforms to this profile where, for its declared enforcement scope, it:¶
establishes the validated credential context before any credential-derived value becomes a decision input (Section 3);¶
evaluates each action against both the credential authority and the current effective authority (Section 5.1);¶
establishes the Mission reference from the credential, or verifies an externally established one where no resource requirement for Mission-bound tokens applies (Section 3);¶
observes Mission state through sources from Section 5.2 that satisfy the runtime core's freshness requirements it adopts for each action class; and¶
honors any classification floor it holds from protected resource metadata (Section 6).¶
It also conforms to the runtime core ([I-D.draft-mcguinness-mission-runtime]) and the issuance profile ([I-D.draft-mcguinness-oauth-mission]) for the same enforcement scope. This profile adds no separate conformance tier: adopting it is what makes "the Mission-bound credential is an OAuth access token" a true statement of the deployment's Enforcement Scope Statement, rather than a second scope to declare separately.¶
The runtime core's Security Considerations ([I-D.draft-mcguinness-mission-runtime]) apply in full, including the remote decision-channel requirement on a PEP/PDP boundary that is not co-resident. General OAuth security guidance [RFC9700] applies to the credentials this profile validates. A PDP that accepts an access token directly, rather than the minimum credential-derived claims a PEP needs to convey, MUST treat it as a credential, protect it against disclosure, and MUST NOT use it outside the declared enforcement scope.¶
Selecting the boundary from aud, rather than checking aud against
a boundary established from routing or configuration, would let a
token with several audiences choose where it is evaluated, and let
the order of its members change the permit binding and the evidence.
An alias taken from the request, or one alias naming two resources,
does the same; Section 3.1 refuses both.¶
Evaluating an action against the approved Authority Set in place of the credential authority would let a narrowed token act with its Mission's broader authority, undoing the narrowing the issuance profile applied; Section 5.1 requires both bounds for that reason.¶
This document defines no evidence content of its own; the runtime core's privacy guidance governs the claims this profile maps into decision inputs and evidence records ([I-D.draft-mcguinness-mission-runtime]).¶
This document registers the following in the "OAuth Protected Resource Metadata" registry ([RFC9728]):¶
Metadata Name: mission_action_class_floors¶
Metadata Description: JSON object mapping a protected resource's action identifiers to minimum runtime action classes.¶
Change Controller: IETF¶
The Mission-bound token claims this document maps are registered by [I-D.draft-mcguinness-oauth-mission].¶
[[ To be removed from the final specification ]]¶
Mission State Sources follows the runtime core's Credential-lifetime freshness rule: token-lifetime expiry and state-gated refresh conform where every issuance, refresh, and exchange path checks Mission state when it mints, and a token minted from an earlier observation adds that observation's age to its lifetime, within the published staleness bound (#1128).¶
The sender-constraint, audience and exp cap citations name the
OAuth binding's Mission-Bound Token Issuance chapter, the retitled
Mission-Bound Access Tokens. No requirement changed (#1892).¶
Boundary and Audience: the runtime core's boundary identifier is the
protected resource's resource identifier, spelled as its RFC 9728
resource value, which trusted static configuration can supply.
aud, a string or an array, names it directly or through an
audience alias, trusted configuration scoped to the validated
issuer and token profile; an array is checked for membership, never
by order. An absent, malformed, or nonmatching audience, or an
ambiguous alias, is refused before any decision, and a stricter
token profile keeps its rule. The input mapping names the boundary
identifier and no longer calls the protected resource the target of
the action. Adds RFC 7519 as a normative reference (#853).¶
Credential Authority and Current Effective Authority: a
mission_resource_access entry's resource matches the operation
target the PEP establishes from its trusted operation contract,
under Resource Access's Resource Matching rules; the token's
audience remains a separate check (#1276).¶
Runtime Input Mapping and Mission State Sources: a state source
reports an expired Mission as terminated with reason expired,
and the Status List cannot provide the termination reason or its
references (#705).¶
Establishing Validated Credential Context: the OAuth realization of
the runtime core's sender-constraint requirement for high-consequence
actions: the cnf binding of the validated credential context (a
JWT's claim, or an opaque token's introspection response), with a
proof the PEP verified; a context with no binding is a bearer token.
Adds RFC 8705 and RFC 9449 as normative references.¶
This document extracts the OAuth-specific realization of Mission-Bound Runtime Enforcement so that document can state a binding-neutral contract. The author thanks reviewers of the runtime core for pressing on the substrate-neutrality claim until it was true. It builds on the OAuth 2.0 Rich Authorization Requests and JWT access token specifications.¶