| Internet-Draft | Outerprise Examination Manual | November 2026 |
| Jones | Expires 21 May 2027 | [Page] |
This document describes repeatable procedures for examining detached evidence packets and the signed findings returned by recipients. It identifies the object, criterion, evidence, procedure, observed result and limits of each finding. The procedures cover artifact identity, release consistency, consent and access, delivery events, benchmark comparisons, commitment openings and policy decisions. Synthetic examples illustrate reproducible checks. The document adapts an examination manual into an implementation-neutral method.¶
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.¶
A detached packet is produced outside the recipient system of record and delivered without requiring a standing connection to that system. Its transport can be email, download, upload or an API. This method helps a reviewer identify which claims a supplied packet supports and which require further evidence. The recipient retains responsibility for its own intake and policy decisions.¶
Use the procedures to examine an explicitly identified system, packet, version and period. Applicability follows the supplied schema, trust configuration and agreed criterion. Product-specific checkpoints and financial workflows are examples. This document defines an examination method and no wire protocol, certification scheme or SCITT conformance profile.¶
The historical source is the examination edition of Surprise, Outerprise [BOOK]. Its 26 preparation questions map to 17 cards on pages 74-90. The method here generalizes those cards and incorporates the comparison-attestation distinctions in the author-supplied revision -01 of [HANDOFF]. The public availability of that revision is a separate repository status question.¶
An artifact is an identified sequence of bytes. A packet consists of artifacts and associated verification material. A content digest identifies bytes produced by a stated serialization recipe. A file fingerprint identifies the literal delivered file. A signature check evaluates identified bytes against a stated key and algorithm. Authority requires an appropriate trust association between key, actor and role.¶
A finding records the result of a named procedure against a stated criterion. A policy decision records an authorized actor’s disposition under an identified policy. A registration receipt concerns registration under a transparency service trust configuration. An acknowledgment concerns a recipient event. Each assertion is scoped to its own evidence.¶
Classify evidence as direct observation, inspected implementation, recorded prior result, or proposed procedure. Preserve the date, origin and limitations of each item. A prior source inspection supports a claim about that source revision. A deployment claim requires evidence about the deployed version.¶
Before testing, identify the asset owner, reviewers, permitted environment, purposes, evidence access and applicable data-handling conditions. Public document review can begin with supplied public artifacts. Active access and mutation tests require authority over the specific environment and records. Prefer synthetic records for controlled tests.¶
For each claim, record a claim identifier, object identifier and digest, source and schema versions, criterion, input evidence references, reviewer, procedure and time, expected result, observed result, conclusion and limitations. Include tool versions, canonical preimage bytes or a protected reference to them, and relevant trust configuration when reproduction depends on them.¶
PASS means the criterion was met within the recorded procedure and scope. FAIL means the procedure produced evidence contrary to that criterion. NOT_TESTED means the necessary procedure was not performed or evidence was unavailable. OUT_OF_SCOPE means the plan excluded that claim. These procedure outcomes apply to individual checks. An NA category outcome belongs to a comparison and has a different role.¶
Preserve an original finding, correction reference, affected versions and retest result. Retesting a later version creates a new result tied to that version. Record partial, interrupted and failed delivery attempts when evaluating operational claims.¶
Identify the eligible row set, declared aggregation and display rule, generated cover and release caller. Recalculate the result using the original inputs. With synthetic scores 10, 20 and 30, the mean is 20.0. Change only the cover to 20.1 and test the ordinary release path. Record the helper response, persisted artifact and delivery outcome. A helper refusal requires caller-path evidence before it supports a claim about release enforcement.¶
Identify the numeric domain and rounding convention. For nineteen scores of 15 and one of 8, the sum is 293 and the exact mean is 14.65. A positive half-up one-decimal display is 14.7. Preserve exact intermediate values. If negative values are supported, state and test their rounding rule explicitly.¶
Identify UTF-8 encoding, key and array order, null handling, excluded fields and timestamp normalization. Reconstruct exact preimages under the artifact’s declared recipe. Change one governed value and check the original reference. Independently hash the literal downloaded file. [RFC8785] supplies one JSON canonicalization scheme. An implementation declaring another recipe requires its own exact specification and test vectors.¶
Identify the designated completion checkpoint and protected fields. Follow a synthetic record through finalization, then attempt approved mutations through each in-scope role path. Record refusal, versioning and any corrected object. Preserve the original version and the relationship to its correction. State the object and field scope of any immutability finding.¶
Trace authorization and consent metadata to the exact purpose and version. Test required metadata through the actual backend entry point and inspect resulting state. Use two authorized synthetic organizations to check own-object access and cross-organization denial. Test public and expired sessions independently. Identify each principal, object, request, response and stored effect.¶
Use synthetic direct identifiers and identifying free text. Inspect collection, participant output, retained storage, organizational projection and delivered packet for each authorized actor. Identify lifecycle-dependent clearing and residual disclosure risk. A digest of a low-entropy identifier can remain guessable. A documented retention or deletion claim requires observations of the relevant lifecycle stage.¶
Vary instrument, formula, population and codebook identity one at a time. Evaluate the declared comparison policy for each change. Identify participant frame, count units, recruitment, exclusions and interval. Preserve a series break when the implemented policy requires it. A permitted numerical comparison supports only its stated population and measurement context.¶
Identify the implemented allowed and denied branches, actor authorization and transition evidence. Test both a valid synthetic sequence and a skipped or denied transition. For actual received-funds claims, reconcile authorized financial records separately. A stored state or digest alone provides no independent evidence of a cash transfer. Domain-specific legal and accounting conclusions require their own criteria and qualified review.¶
Reconstruct paired organization identifiers under the declared schema. Record corrected identities and retained versions. Examine a contract or execution reference as a separate evidence object. A relationship identifier supports identification within its schema. Signer identity, authority and execution each require evidence appropriate to that claim.¶
Record generation, release, provider acceptance, download, recipient acknowledgment and recipient action individually. Compare delivered file bytes with an authenticated expected fingerprint. Test retry and duplicate handling within scope. Attribute each timestamp to its clock and observer. Report measured intervals and failures with their defined start and end events.¶
For each comparable route X, calculate K_X = C_X + T_X * V. C_X is a non-negative integer paid cost in a stated currency’s minor units. T_X is non-negative whole elapsed calendar days to the identified accepted deliverable. V is a non-negative minor-unit value per day declared by the recipient before calculation, with zero used when none is declared. Compare the same deliverable, scope and period.¶
The signed difference is K_E - K_O. Compute the multiple K_E / K_O only when K_O is positive. Round the display half up to two decimal places after the exact calculation. Preserve zero and negative differences. Identify the evidence source and recorded, modeled or missing classification for every input.¶
With V = 0, 800000 minor units for an observed current route and 500000 for a modeled unused route produce a modeled difference of 300000 and a multiple of 1.60. An observed-only baseline requires observed evidence for both routes. Distinguish paid cash costs and monetized elapsed time when interpreting the result.¶
Pin the request and packet digests, benchmark identity and source, currency, period, scope, intended audience and non-overlapping category definitions. For each category report GT, EQ or LT when corresponding recipient and benchmark amounts are comparable. Calculate the total from complete sums. The total cannot be inferred from category labels alone.¶
Where a category cannot be compared, report NA with the signed reason UNAVAILABLE, NOT_APPLICABLE or NOT_ASSESSED. Preserve every category. Any NA makes the total NA and identifies the affected categories. An unopened commitment alone supplies no reason to change a category to NA.¶
Report commitment opening as NO_COMMITMENT, NOT_PROVIDED, OPENED_MATCH or OPENED_MISMATCH. A matching opening binds disclosed figures to the commitment. Independent reconciliation to source records requires separate evidence. If a separate proof is supplied, identify the specific claim and verification procedure. Report only the claim the proof establishes.¶
For the illustrative commitment in [HANDOFF], hash ASCII DEH-COMMIT-1, one zero octet, the fresh secret 16-byte salt, then canonical UTF-8 JSON containing decimal-string figures. Repeated thresholds and disclosed category labels may reveal private information. The public synthetic salt exists for reproduction and is unsuitable for private production figures.¶
A separate policy record identifies the exact attestation digest, NA findings considered, policy identifier and version, decision-maker, scope, conditions and validity interval. Verify its signature and decision-maker authority separately. Record any permitted exception and basis when complete evidence is normally required. ALLOW leaves evidence findings as recorded. This method specifies no subject-based supersession mechanism or extensible reason registry.¶
The following example illustrates reporting only. No signature was executed for the example. Its signature result is assumed. Real findings require the corresponding checks against actual supplied bytes, keys, trust configuration and source evidence.¶
claim: DEMO-003 (synthetic) attestation signature: PASS against stated key (assumed) signer authority: not established by signature check commitment opening: NOT_PROVIDED independent verification of figures: NOT_ESTABLISHED integration: GT middleware: NA (UNAVAILABLE) decommission: LT total: NA (middleware) policy record: ALLOW for a bounded pilot (illustrative) policy signature and authority: separate checks required¶
The author locally checked the arithmetic cases and disclosed commitment listed here when preparing this revision. These checks exercise synthetic inputs. They do not constitute an independent audit, a source-ledger reconciliation, a deployed authorization test or a signature verification run.¶
For benchmark amounts 100000, 50000 and 20000 and recipient amounts 130000, 50000 and 15000, the outcomes are GT, EQ and LT. The complete totals are 195000 and 170000, giving GT. When middleware is NA (UNAVAILABLE), the total is NA. The comparison procedure preserves the signed reason.¶
For the commitment example, the salt is hexadecimal 000102030405060708090a0b0c0d0e0f. The canonical JSON is {"decommission":"15000","integration":"130000","middleware":"50000"}. The expected SHA-256 is 36d516c50e28dabc4cfc12298fd75046b6dc10dae2a9bc873d1c3a4cb3494ef2. Changing integration to 130001 produces a different commitment.¶
Additional arithmetic cases use equal costs, a negative difference, a zero denominator and positive half-up rounding. For K_E = 201 and K_O = 200, the two-decimal displayed multiple is 1.01. Each calculation uses integer or exact decimal arithmetic.¶
A trusted reference for expected digests is necessary to evaluate integrity. An attacker able to replace both payload and reference can preserve apparent consistency. Signature verification depends on algorithm handling, exact bytes and a trusted association between the key and the claimed actor or role. Record verification errors and unsupported algorithms explicitly.¶
Constrain active tests to authorized assets and synthetic data. Controlled negative tests can alter stored state or trigger delivery, so record expected effects, stop conditions and restoration. Inspect the complete permitted release paths and tenant boundaries within scope. Producer, delivery provider and recipient can have distinct trust and failure boundaries.¶
Identify the expected request, audience, period, scope and packet when evaluating a return record. Apply the recipient’s explicit rules for freshness, replay, duplicates, supersession and exceptions. Preserve findings about an earlier version when later evidence changes the disposition.¶
Minimize identifiers, raw participant content and confidential source figures in public findings. Share protected reconstruction inputs only under the relevant purpose and authority. Public logs and repeated comparison requests can create inference and cross-context correlation risks. Consent and data-access rights require their own evidence.¶
A stored hash of a small input domain can support offline guessing. Keep production commitment salts fresh and secret under the declared scheme. Document what a recipient learns from a comparison result and whether repeated benchmarks narrow private values.¶
This document requests no IANA actions.¶
The source book card numbers are retained here for traceability. Its historical implementation findings remain dated source claims. The generic procedures in this document can be applied to another implementation under that implementation’s own schema and scope.¶
Cards 01-03 address release consistency, rounding and caller behavior. Cards 04-05 address digest inputs and reconstruction. Cards 06-07 address finalization and consent. Card 08 addresses version comparability. Cards 09-10 address privacy and role isolation. Cards 11-12 address approved and denied transitions. Card 13 addresses paired identities and execution. Card 14 addresses retained bytes. Card 15 addresses financial receipt. Card 16 addresses delivery and recipient events. Card 17 addresses professional scope and evidence limits.¶
Addendum B to the book, dated November 17, 2026, adds Cards 18-23 for route-cost inputs, category outcomes and NA, commitments, authority and policy binding, EOT/EOI and version continuity. The complete new addendum accompanies the revised book. The public BOOK reference identifies the earlier source edition.¶
The following 21 company vocabulary entries are informative examples derived from [LEXICON] and the vocabulary appendix of [HANDOFF]. Terms 1-16 are covered by the earlier book and its Addendum A. Terms 17-21 complete the Version 4 inventory. The app snapshot observed on November 17 identified SEALED Version 4.¶