The Continuity Evaluation Kit

Continuity Is Not One Thing argues that delegated work poses four independently governed continuity questions, and it was written to age well: roles instead of versions and statuses, mechanisms named only where they prove a distinction, and the model stated once. This page is its working companion, and it makes the opposite trade. Everything here is built to be revised: the vocabulary that keeps the essay’s terms honest, the rubric for evaluating proposals the essay never mentions, the worked demonstrations, and the standards map with the versions and statuses that change faster than architecture. Read the essay for the model. Use this page at the review meeting.

The Objects

Continuity arguments slide when one word carries several objects. Six terms keep them apart:

The separations do immediate work. A Transaction Token is an artifact carrying identity and authorization context; it is not the authority for either. An identity assertion is evidence, not permission. A token exchange is a protocol operation, not a semantic guarantee about what its output means. A Mission is state, which artifacts may reference but cannot be. And sender constraint validates an artifact’s presentation, never the truth of the facts inside it. A joining authority may associate facts it does not own; that entitlement is scoped, never ambient.

The Rubric

Any continuity proposal, draft, profile, or product feature can be evaluated with twelve questions. None of them requires accepting the four-question model, any particular record format, or any proposal of mine; they only require the proposal to say what it establishes and for whom.

A proposal that answers all twelve is evaluable whether or not it is good. A proposal that cannot answer numbers 1 and 2 is not yet a security mechanism; it is a data structure.

The rubric is the essay’s eight-element boundary contract in evaluation form:

Question 12 is the deployment assumption beneath all eight.

Two Worked Evaluations

The rubric earns its keep on real work, so here it is applied to the two most instructive current drafts: one intra-domain propagation mechanism and one cross-domain re-issuance pattern. The point is not the verdicts, both drafts fare well, but that the rubric surfaces exactly where each one’s obligations end.

Two lessons generalize. Transaction Tokens answer every question inside one trust domain and go silent at its wall, which is not a defect but a scope statement the rubric makes visible. Identity Chaining answers the cross-domain questions precisely because it refuses to carry target-local authority, leaving the target’s decision where the essay’s model says it belongs.

The Working-Group Queue, Classified

The framing is testable against the official queue: if the taxonomy is right, every active OAuth working-group document should classify without forcing a fifth question, and the gaps the essay names should appear as absences. The full set of sixteen active documents, as of August 10, 2026:

The test returns three results. Nothing in the queue breaks the taxonomy or demands a fifth question, and only three documents are answers to continuity questions; the rest classify as presenter binding, admission, carriers, lifecycle elements, and substrate, the categories the essay defines around the questions, with two documents standardizing contract elements that involve no boundary crossing at all. The queue’s center of mass is the precondition: three documents harden who presents, and the client, workload, and actor reconciliation gap is visible inside that cluster, since the workload joins to the client while actor evidence binds to neither. And the essay’s gap list appears as an absence pattern: request continuity exists only as an individual draft, while work continuity, joining authorities, comparison vocabularies, and derivation-graph revocation have no working-group document at all. Rerun this table each IETF cycle; a document that resists classification, or a new adoption inside a named gap, is exactly the correction this page exists to record.

The Map, With Dates

The essay’s map records each proposal’s primary role only. This one carries the churn, and it is dated so its staleness is measurable: statuses below are as of August 8, 2026.

The deployed substrate everything composes over:

The emerging continuity work, with the tracking detail the essay refuses to carry. The entries from my own program, the Identity Continuation Assertion, the Actor Profile with its proofs and receipts, the trust-framework, domain-authorized-issuer, and instance-identity drafts, the design sketches, and Mission-Bound Authorization, are disclosed here as in the essay:

The gaps themselves, what no row supplies, live in the essay’s work list, where they belong to the argument rather than to the tracking. And this map is scoped to the four questions; the Mission-relationship view of the same territory, the full WIMSE and OpenID inventories, the 2026 agent-draft wave, and the substitution hazards, lives in the published standards map, reconciled on its own date.

Using the Kit

At a design review, walk the essay’s boundary contract at each boundary a change touches, then run the rubric here against whatever mechanism is proposed to fill its elements. At a standards review, the rubric is the comment template: a draft that cannot say what fact it establishes, for which verifier, with what failure behavior, is not ready for the argument about whether its answer is right. And when the map above disagrees with the datatracker, trust the datatracker and tell me; this page expects to be corrected, which is exactly the property the essay was built not to need.