SRA Modeled-vs-Real Manifest

Demonstrated for real

Warrant signing and gate verification. The page generates an ephemeral Web Crypto keypair, canonicalizes the immutable outward warrant payload, signs it, and verifies the signature against the known public key during gate step 4. The displayed prefix is ed25519: or ecdsa-p256: according to the algorithm actually used.

Modeled and labeled

Key custody. UI label: “Keys are generated in-page and ephemeral, demonstrating the signature/verification mechanism. Production requires key custody in an HSM/KMS under community control.”

Not attempted, and why

HSM/KMS custody, key rotation, and compromise runbooks. A public static page cannot provide community-controlled hardware custody, durable issuer operations, or incident response.

Demonstrated for real

JSON canonicalization. A faithful RFC 8785/JCS subset sorts object keys, preserves array order, rejects non-JSON values and invalid Unicode scalar sequences, and uses ECMAScript number serialization before signing and verification.

Modeled and labeled

Governance legitimacy. UI label: “A signature proves who signed bytes; it does not prove that the governance process was legitimate or that the claims are true.”

Not attempted, and why

Authentication of governing authorities. The browser has no authoritative identity provider or community governance trust root. Signing a fixture cannot establish standing.

Demonstrated for real

CSPRNG identifiers. Warrant, determination, gate-decision, contest, capability, public-statement, restricted-note, translation, issuer, profile, evidence, and resource identifiers use crypto.getRandomValues; no public and hidden sequence is shared.

Modeled and labeled

Principal authentication. Gate label: “Assume authenticated principal synthetic-principal. No identity provider is contacted, and authentication is not standing.”

Not attempted, and why

OIDC, institutional identity federation, and standing adjudication. Authentication can identify a principal; it cannot independently confer warrant-issuing authority.

Demonstrated for real

DVA issuance without a computed verdict. The condition worksheet displays considerations without numbers or weights. Each proposed operation remains PENDING_GOVERNANCE until governance enters its determination; no aggregate verdict is recorded. The browser validates proposal scope, profile scope, issuance attestation, signature, and exact operation coverage; it never chooses a viability outcome.

Modeled and labeled

Synthetic governance entry. UI label: “A governance determination has been entered under the synthetic profile's issuance rule.”

Not attempted, and why

A real DVA convening process. The browser cannot determine who has standing, whose witnessing is required, how dissent is carried, or what must remain unnamed under community law.

Demonstrated for real

Distinct object stores. DvaDetermination, GateDecision, ContestedDesignation, PublicRefusalStatement, and RestrictedGovernanceNote occupy separate in-memory structures. A gate denial and a non-viability determination write no refusal statement.

Modeled and labeled

Store separation. UI label: “This separate browser-memory structure demonstrates separation only. It is not access control or secure storage.”

Not attempted, and why

Separated production databases, service accounts, keys, and backups. In-memory maps demonstrate type and write-path separation, not operational isolation or authorization.

Demonstrated for real

NDE non-recording at the shared registry boundary. There is no NDE tier, counter, object, registry entry, count, or placeholder in the public and restricted stores. The two external-query fixtures return identical status, body, headers, and byte length.

Modeled and labeled

NDE response shape only. UI label: “In a real deployment, equivalence across timing, caching, rate-limiting, logs, and backups must be measured under adversarial testing; it cannot be inferred from code. This model demonstrates the response shape only.”

Not attempted, and why

Statistical NDE timing-equivalence testing. Meaningful evidence requires a production-like service, caches, load balancers, logs, backups, rate limits, traffic generation, and adversarial measurement.

Demonstrated for real

Constraint-preserving receiver compatibility. Typed rules cover permissions, prohibitions, affirmative obligations, temporal conditions, audience boundaries, derivative rules, and retention/deletion. Exact preservation returns COMPATIBLE; a real signed, versioned policy translation can return AUTHORIZED_TRANSLATION; unilateral weakening or tightening returns INCOMPATIBLE; unknown terms return INDETERMINATE. INDETERMINATE never releases.

Modeled and labeled

Receiver declaration. The receiver profile is a synthetic declaration of accepted policy and supported obligations. A passing browser check does not show that an institution will obey after receiving plaintext.

Not attempted, and why

Receiver federation and attestation. The static model cannot establish institutional trust, inspect remote configuration, compel legal compliance, or prevent screenshots, copying, exfiltration, and bad-faith reuse.

Demonstrated for real

Operation-scoped contest behavior. A signed ContestedDesignation contains opaque identifiers, scope, interim rule, process reference, activation and review times, and state sequence. Other gate checks still run. Irreversible disputed disclosure fails closed; an uncontested storage operation may continue. Passing review_after requires re-attestation and never becomes consent. Resolution requires a different, newly signed warrant.

Modeled and labeled

Recognized process. The fixture uses the synthetic governance profile as the recognized process boundary. An unverified outsider assertion is ignored and cannot freeze a resource.

Not attempted, and why

Governance adjudication. The model protects the interval but cannot decide which witnessing is faithful, resolve disagreement, or make a captured process legitimate.

Demonstrated for real

Fail-closed gate order. The gate performs the twelve ordered checks: principal stub; opaque handle; latest warrant; real signature and trusted issuer; warrant state; operation-scoped contest; exact scope; receiver compatibility; community consult stub; obligations; capability; synthetic reveal. Unknown policy, malformed dates, stale revocation, unsupported obligations, and indeterminate compatibility deny.

Modeled and labeled

Community-decision-service consult. Gate label: “Modeled community-decision-service consult: ALLOW. No real community decision service is contacted.”

Not attempted, and why

A real community decision service. Protected context, availability, authenticated requests, revocation, and community-controlled operations require a service outside the static page.

Demonstrated for real

Affirmative-obligation checks. Known obligation types are validated against receiver support. A due obligation must carry opaque evidence; a past-due obligation without evidence denies.

Modeled and labeled

Evidence meaning. Opaque fixture references demonstrate the field and due-date rule, not the truth, sufficiency, or community acceptance of evidence.

Not attempted, and why

Audit and evidence verification. The browser has no trusted evidence store, reviewer process, or durable audit trail.

Demonstrated for real

Capability lifecycle and scope. The model mints a real signed, short-lived capability only after the preceding checks pass. Consumption verifies signature, expiry, exact resource/action/audience/workflow/purpose, and an in-memory single-use set. Reuse has no second side effect. Only then is a synthetic payload revealed.

Modeled and labeled

Transport boundary. UI label: “Sender-constraining (mTLS/DPoP) and atomic key release require a real service; this model demonstrates the token lifecycle and scope-checking.”

Not attempted, and why

mTLS/DPoP, replay-resistant distributed consumption, and atomic key release. These require a sender, receiver, network protocol, durable atomic state, and protected key material.

Demonstrated for real

Deliberate refusal authorship. Public and restricted records require an explicit type, governance-authored text, profile scope, attestation, real signature, CSPRNG identifier, and idempotency key. No prose is generated from gate diagnostics, and an absent or invalid disclosure posture creates no record.

Modeled and labeled

Restricted visibility. The model labels the restricted store as an in-memory structural demonstration, not an externally enforced audience boundary.

Not attempted, and why

Secure restricted discovery and authorization. A public browser cannot provide partner authentication, authorization, compartmented queries, or durable retention controls.

Demonstrated for real

Strict browser-side schemas and safer rendering. Unknown fields and enums fail closed in the decision model. User-entered text is rendered with DOM text nodes, not HTML parsing. The SRA routes use an external-script/external-style Content Security Policy without unsafe-inline.

Modeled and labeled

Browser accessibility checks. Native controls, associated labels and fieldsets, keyboard tabs, visible focus, text status labels, aria-live, reduced-motion handling, and narrow-width reflow model an accessible interaction surface.

Not attempted, and why

Production security and full assistive-technology certification. A static review cannot substitute for independent penetration testing, multiple browser/AT combinations, organizational accessibility testing, or ongoing regression monitoring.

Demonstrated for real

No sensitive form persistence. Proposal and authorship drafts remain in memory, are never written to URLs or localStorage, and are reset on page exit.

Modeled and labeled

Status-banner preference. A session-only collapsed/expanded preference contains no form or protected state.

Not attempted, and why

Storage of real protected data. Explicitly prohibited. This model is not a repository, record system, or secure processing environment.