Review · November 10, 2026
CIMD Software Statement Lifecycle, Withdrawing and Replacing a Third-Party Review After the Grant Is Open
The problem it's solving
The CIMD Software Statement draft lets a third party sign off on one exact version of a client's Client ID Metadata Document. The AS checks that statement once, when the grant opens. Refresh checks the client's key and nothing else.
That leaves a gap. An issuer that finds a defect, or approves less than it did before, has no way to make existing grants follow the new review. Old statements stay acceptable until they expire, and expiry ends no grant.
This draft fills that gap with three things: an issuer can withdraw a statement early, a newer statement displaces older ones for new grants, and a refresh can be made to depend on a current statement.
How it fits in the working group
It fits the charter, but only as an extension of two individual drafts by the same author. Neither the base statement draft nor the registration companion is adopted. This document has no meaning without them.
It does not add a role. The statement issuer picks up two new jobs: publishing a status list and, optionally, running a Shared Signals transmitter. The AS becomes a status list consumer and, optionally, a Shared Signals receiver for each issuer it trusts.
The normative dependencies are the two companion drafts, Token Status List, the OpenID Shared Signals Framework, RFC 6749, RFC 8414, and four Security Event Token RFCs.
The wire surface is small: one AS metadata value, one event type URN, and one SET payload claim. Most of the document is AS-side state and processing rules.
What the mechanism actually does
Withdrawal produces a "refusal record." The issuer flips the statement's entry in a Token Status List. The AS learns this the next time it fetches the list, then refuses new admissions, registrations, and refreshes under that statement. An AS operator can also enter a refusal by hand.
Withdrawal revokes nothing. Issued access tokens keep working. The grant is not ended. It is blocked at the next refresh until the client supplies a newer statement. If the issuer later restores the status, the grant can refresh again.
Supersession uses an "iat floor." The AS remembers the latest iat it has accepted for each issuer, client URL, and use. A statement with an earlier iat can no longer open anything new. Grants already opened on the older statement continue.
Refresh can require a replacement. A replacement is required when the recorded statement is withdrawn. For an expired statement, the AS SHOULD require one. The client sends the newer statement in the software_statement parameter on the refresh request, or the AS pulls it from the client's published list.
The replacement must match the live document. It needs the same issuer and client URL, a later iat, and a digest of the CIMD currently served. That document must still carry the key the grant is bound to.
An optional Shared Signals event speeds things up. The issuer sends a "status changed" SET naming the client URL. The event carries no status. It only tells the AS to fetch the status list now instead of on its schedule.
The issuer runs three separate key sets. One for statements, one for status list tokens, one for events. They must be at three locations with no shared keys.
Significant interoperability concerns
The same grant behaves differently at every AS. Consulting the withdrawal mechanism is a SHOULD. Requiring a current statement after expiry is a SHOULD. What to do when the status list is unreachable is local policy. No withdrawal mechanism is required at all. An issuer that withdraws a statement cannot know which servers will act on it or when.
Supersession only works where the newer statement has been seen. The floor advances when an AS accepts a later statement. An AS that never sees it keeps accepting the older, broader one. The draft says so in Section 12.2 and tells issuers to also withdraw. Withdrawal is optional for both sides.
The client finds out at refresh time. Nothing tells a client its statement was withdrawn or expired. The refresh fails with statement_required. That one code covers a withdrawn statement, a superseded one, a failed replacement, and a required issuer mismatch where no replacement can help. The client can't tell whether to fetch a new statement, wait for re-review, or send the user through authorization again.
There is still no way to get a new statement. Issuance is out of scope across all three drafts. Refresh replacement depends on the client holding a statement newer than the one the AS recorded. Every client and issuer pair has to invent that delivery path.
The byte-exact digest problem now reaches refresh. A replacement must match the document currently served. A client that edits its CIMD before the issuer re-reviews it can't replace its statement. If the old statement is expired or withdrawn at that point, every refresh fails until the issuer catches up.
The metadata doesn't answer what a client needs to know. software_statement_replacement_supported says the AS will apply a replacement. It doesn't say whether the AS requires a current statement on refresh, or whether it checks status at all.
It overrides Token Status List key resolution. Status list tokens are verified with the issuer's RFC 8414 jwks_uri or a configured URL, never with keys the token points to. An issuer's existing status list deployment may not match.
Mechanisms that could be removed
- The Shared Signals event (Section 7). It carries no status and changes no decision. It is a cache-invalidation hint that requires the AS to run an SSF receiver and a stream per issuer. A shorter
ttlon the status list does the same job. This removes five pages, a third key set, and the dependency on SSF, CAEP, and four SET RFCs. - Per-tenant and per-use iat floors. Separate floors for presentation and registration, per tenant and server-wide, with a special case for unrestricted registrations. This is AS-internal bookkeeping. One rule would do: don't accept a statement older than the newest one you've accepted for that issuer and client.
- Operator-entered refusal records. An AS operator can always block a client. That needs no specification.
- Status list rollback tracking. Keeping the last
iatand a SHA-256 of each list token belongs in Token Status List if it's needed. - Table 2 and Section 8. Nine rows of ordered error mapping that mostly resolve to
statement_required. - Sections 10 and 11. Both open by saying they add no requirement.
There is a larger question behind this list. Section 11.1 tells issuers to choose exp as if no AS uses a withdrawal mechanism. If lifetimes are short and refresh requires a current statement, the status list adds little. Making the Section 5.1 rule a MUST and dropping Sections 6 and 7 would give predictable behavior in about eight pages.
Writing quality
The writing is precise and in correct IETF style. Table 1 is the best part of the document. It shows in four rows what each statement state allows.
It is hard to read alone. Nearly every paragraph cites a numbered step or section of the two companion drafts. "Establishment," "Proven Key," "Selected Method," "Effective Expiry," and "Governing Statement" are all defined elsewhere. This draft adds five more terms.
At 29 pages it is long for what it changes on the wire. The three-document family now runs past 100 pages for one JWT claim and its consequences.
Some rules are stated three times. The refresh error mapping appears in Section 5.2 prose, Table 2, and Section 8.
One small issue. It cites both companions at -00. Both were at -01 on the same day.
Background of the author
Sole author, Karl McGuinness, listed as Independent, formerly at Okta.
The Datatracker profile could not be fetched for this review. The figures below are from the profile as checked for the October reviews, plus what is in the drafts archive.
- 0 published RFCs
- 16 active Internet-Drafts and 2 expired, as of mid-October 2026
- Co-author of ID-JAG, an adopted WG document
Not a new author, and writing at a very high rate. The archive holds 21 distinct individual OAuth drafts under his name. This draft went from -00 to -01 on 2 November, the same day as CIMD -03.
Questions worth raising
- Should this be a separate document? The base draft checks a statement once and lets grants outlive it indefinitely. Sections 4 and 5 here are the fix. A reviewer can't judge the base draft's security without them.
- Is a status list needed if refresh requires a current statement? Short lifetimes plus a mandatory re-check bound the exposure without any issuer infrastructure. The draft itself tells issuers not to count on withdrawal.
- Why does withdrawal block refresh but leave the grant alive? A defect serious enough to withdraw a review may warrant revoking the refresh token. The draft leaves that to the operator entering refusal records.
- How does a client learn it needs a new statement before its refresh fails? Without an issuance and notification path, the first signal is a production error.
- Is the Shared Signals event worth an SSF dependency in an OAuth WG document? It prompts a fetch the AS would make anyway.
For the two drafts this one depends on, read the base statement draft and the registration companion.