You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
DateKeys/testdata/README.md

509 lines
27 KiB

# DateKeys test data
Official vectors, fixtures and corpora of the DateKeys Protocol Specification
Implement capsule format 2 of spec v0.9 The reference moves to the DateKeys Protocol Specification v0.9, approved by its author on 29 September 2026. Encrypt writes capsule format 2 only; Open and Inspect read formats 1 and 2, and a format 1 capsule keeps the verdict v0.8.2 gave it. Format 2 (spec §22, §29.1, §31, §39): - VERSION in the PRELUDE is the capsule format, capsule.Format; any other value is ERR_UNSUPPORTED_VERSION at step 2. - CONTROL_CBOR has the schema version of its format. Version 2 adds key 6, payload_length (8 bytes, big-endian, at most L_MAX = 2^53 - 2^46), and key 7, padding (1 bloque256, 2 reforzado); it is 103 bytes without extensions, whatever L. - The payload is the content padded with zeros to P = rule(L). Step 17 checks the length and the zeros, and Open writes only the first L bytes. - INNER_ACCESS_AGE holds exactly 16 X25519 stanzas: 1 to 16 credentials, and a dummy in each slot left, in a uniformly random order. Writer rules (spec §62.1): EncryptOptions.Length is required and the source must deliver exactly that many bytes; recipients that are not canonical or of low order are rejected (agewrap.CheckX25519Recipient); self-checks of the header, the control, INNER_ACCESS_AGE and PAYLOAD_AGE. The CLI measures its input, takes -padding and reports the format. Test data: seven format 2 fixtures, padding vectors checked against math/big, format 2 CBOR vectors, and the mutation corpus in both formats with the 22 cases of the third list of spec §64, built without randomness by sealing the fixtures again with their known keys and nonces. The format 1 fixtures are kept byte for byte and never regenerated; the differential corpus keeps its 1825 cases and adds a block per format 2 fixture. The spec copy loses its "to be implemented" markers, and the READMEs, CHANGELOG, traceability and testdata/README.md follow v0.9. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
v0.9, generated by the reference implementation. Another implementation
consumes them as they are: this file documents every format, so that no Go code
Spec v0.8.2 refinements: error precedence, trust model, strict order Approved refinements, each recorded with its reproducible case in the §76 v0.8.2 subsection: - §69.1: layered error model with normative precedence (frame, type tag and version, CBOR profile and CDDL, then fields with their own code in ascending key order; across steps the §63 order decides), with a scope paragraph for the optional steps 5, 6 and 8. - §55.1: normative trust table per section (who can write it, from which step it is bound, what it never proves); §72: security-relevant claims go in CONTROL_CBOR or under a signature, .dkk data is advisory. - §31/§54: extension arrays in strictly ascending unsigned byte order of extension_id (one rule for order and uniqueness). - Gaps a second implementation needed: §28.1 malformed age headers, §15/§19 latest unlock time and dk1_ reading rules, §22/§23/§57 length lower bounds, §63 step 8 tlock argument comparison and step 9 order, §12.1 profile validation with the drand chain-hash formula, §74 table of implementation limits. Reference alignment: .dkk errors only at step 9.a (new OpenOptions.AccessKeyFile, used by the CLI), CR/LF in dk1_ is ERR_DATEKEY_INVALID, BODY_LEN 0 is ERR_INTEGRITY, nil identities are not credentials, and AccessIdentity tries every identity on every stanza so its verdict does not depend on their order. dk1.json gains three vectors; every other testdata file is byte-identical. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks ago
has to be read. The rules that decide each verdict are in the specification;
this file points to them, and states only what belongs to the files
themselves.
```
go run ./internal/testkit/genfixtures -out testdata
```
regenerates everything except the `.dkc` and `.dkk` fixtures, which are
Implement capsule format 2 of spec v0.9 The reference moves to the DateKeys Protocol Specification v0.9, approved by its author on 29 September 2026. Encrypt writes capsule format 2 only; Open and Inspect read formats 1 and 2, and a format 1 capsule keeps the verdict v0.8.2 gave it. Format 2 (spec §22, §29.1, §31, §39): - VERSION in the PRELUDE is the capsule format, capsule.Format; any other value is ERR_UNSUPPORTED_VERSION at step 2. - CONTROL_CBOR has the schema version of its format. Version 2 adds key 6, payload_length (8 bytes, big-endian, at most L_MAX = 2^53 - 2^46), and key 7, padding (1 bloque256, 2 reforzado); it is 103 bytes without extensions, whatever L. - The payload is the content padded with zeros to P = rule(L). Step 17 checks the length and the zeros, and Open writes only the first L bytes. - INNER_ACCESS_AGE holds exactly 16 X25519 stanzas: 1 to 16 credentials, and a dummy in each slot left, in a uniformly random order. Writer rules (spec §62.1): EncryptOptions.Length is required and the source must deliver exactly that many bytes; recipients that are not canonical or of low order are rejected (agewrap.CheckX25519Recipient); self-checks of the header, the control, INNER_ACCESS_AGE and PAYLOAD_AGE. The CLI measures its input, takes -padding and reports the format. Test data: seven format 2 fixtures, padding vectors checked against math/big, format 2 CBOR vectors, and the mutation corpus in both formats with the 22 cases of the third list of spec §64, built without randomness by sealing the fixtures again with their known keys and nonces. The format 1 fixtures are kept byte for byte and never regenerated; the differential corpus keeps its 1825 cases and adds a block per format 2 fixture. The spec copy loses its "to be implemented" markers, and the READMEs, CHANGELOG, traceability and testdata/README.md follow v0.9. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
generated once and frozen (spec §67). The records of each fixture
(`<name>.json`, `<name>.dkk.json`, `<name>.inspect.json`) are recomputed from
its frozen bytes, so the format 1 fixtures of v0.8.2 carry `"spec": "0.9"` and
the fields added since. The local gate (`scripts/check.sh`) and CI run it and
fail if any committed file changes: every file below is exactly what the
implementation computes today.
Conventions for every file:
- JSON in UTF-8, with LF line endings. Binary values are lowercase hex strings.
- `error`, `result` and similar fields hold the normative codes of spec §69, such
as `ERR_NON_CANONICAL_CBOR`.
- `step` is a step of the reading flow of spec §63, 1 to 18. Steps 1 to 8 are the
pre-unlock checks (`datekeys inspect`, `capsule.Inspect`): no network, no
secret.
- The `spec` field names the version of the specification.
Implement capsule format 2 of spec v0.9 The reference moves to the DateKeys Protocol Specification v0.9, approved by its author on 29 September 2026. Encrypt writes capsule format 2 only; Open and Inspect read formats 1 and 2, and a format 1 capsule keeps the verdict v0.8.2 gave it. Format 2 (spec §22, §29.1, §31, §39): - VERSION in the PRELUDE is the capsule format, capsule.Format; any other value is ERR_UNSUPPORTED_VERSION at step 2. - CONTROL_CBOR has the schema version of its format. Version 2 adds key 6, payload_length (8 bytes, big-endian, at most L_MAX = 2^53 - 2^46), and key 7, padding (1 bloque256, 2 reforzado); it is 103 bytes without extensions, whatever L. - The payload is the content padded with zeros to P = rule(L). Step 17 checks the length and the zeros, and Open writes only the first L bytes. - INNER_ACCESS_AGE holds exactly 16 X25519 stanzas: 1 to 16 credentials, and a dummy in each slot left, in a uniformly random order. Writer rules (spec §62.1): EncryptOptions.Length is required and the source must deliver exactly that many bytes; recipients that are not canonical or of low order are rejected (agewrap.CheckX25519Recipient); self-checks of the header, the control, INNER_ACCESS_AGE and PAYLOAD_AGE. The CLI measures its input, takes -padding and reports the format. Test data: seven format 2 fixtures, padding vectors checked against math/big, format 2 CBOR vectors, and the mutation corpus in both formats with the 22 cases of the third list of spec §64, built without randomness by sealing the fixtures again with their known keys and nonces. The format 1 fixtures are kept byte for byte and never regenerated; the differential corpus keeps its 1825 cases and adds a block per format 2 fixture. The spec copy loses its "to be implemented" markers, and the READMEs, CHANGELOG, traceability and testdata/README.md follow v0.9. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
- `format` is a capsule format, the `VERSION` of the PRELUDE (spec §22): 1 or
2. A writer produces format 2 only; format 1, that of v0.8.2, is still read
(spec §70).
| File | Content | Spec |
|---|---|---|
| `vectors/profile_quicknet.json` | Quicknet Provider Profile: its canonical CBOR and `profile_hash` | §11, §12 |
| `vectors/quicknet_rounds.json` | date → round resolution | §15, §16, §65 |
| `vectors/dk1.json` | canonical `dk1_` strings, and rejected encodings with their code | §18, §19, §66 |
Implement capsule format 2 of spec v0.9 The reference moves to the DateKeys Protocol Specification v0.9, approved by its author on 29 September 2026. Encrypt writes capsule format 2 only; Open and Inspect read formats 1 and 2, and a format 1 capsule keeps the verdict v0.8.2 gave it. Format 2 (spec §22, §29.1, §31, §39): - VERSION in the PRELUDE is the capsule format, capsule.Format; any other value is ERR_UNSUPPORTED_VERSION at step 2. - CONTROL_CBOR has the schema version of its format. Version 2 adds key 6, payload_length (8 bytes, big-endian, at most L_MAX = 2^53 - 2^46), and key 7, padding (1 bloque256, 2 reforzado); it is 103 bytes without extensions, whatever L. - The payload is the content padded with zeros to P = rule(L). Step 17 checks the length and the zeros, and Open writes only the first L bytes. - INNER_ACCESS_AGE holds exactly 16 X25519 stanzas: 1 to 16 credentials, and a dummy in each slot left, in a uniformly random order. Writer rules (spec §62.1): EncryptOptions.Length is required and the source must deliver exactly that many bytes; recipients that are not canonical or of low order are rejected (agewrap.CheckX25519Recipient); self-checks of the header, the control, INNER_ACCESS_AGE and PAYLOAD_AGE. The CLI measures its input, takes -padding and reports the format. Test data: seven format 2 fixtures, padding vectors checked against math/big, format 2 CBOR vectors, and the mutation corpus in both formats with the 22 cases of the third list of spec §64, built without randomness by sealing the fixtures again with their known keys and nonces. The format 1 fixtures are kept byte for byte and never regenerated; the differential corpus keeps its 1825 cases and adds a block per format 2 fixture. The spec copy loses its "to be implemented" markers, and the READMEs, CHANGELOG, traceability and testdata/README.md follow v0.9. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
| `vectors/cbor.json` | the CBOR profile, and one block of vectors per schema, CONTROL_CBOR in both formats | §58, CDDL |
Spec v0.8.2: corrections from the formal review A formal review of the whole v0.8.2 text found it approvable after these corrections, recorded in §76 ("Correcciones de la revisión formal"): - §27 no longer calls header_binding the authenticity of PUBLIC_HEADER: it binds the header to the opened control, never authorship or date (§55.1); the age MAC only protects against whoever lacks the file key. - §63 steps 9 and 10: a network source (relay, Release API, cache) MUST verify every response and gives ERR_RELEASE_UNAVAILABLE at step 9 when none verifies; the step-10 codes are for a directly supplied release. The reference already behaved so; TestReleaseFromANetworkSource pins both paths. - §54 and §72: registrations declare the objects and arrays where an extension may appear, and a known extension out of place counts as unknown there. The reference gains the optional extension.Placement interface, used at steps 4, 9.a and 14. - §63 step 11 fixes the GT serialization hashed by H2 (kilic/kyber order) with the frozen vector H2(e(G1, G2))[:16] = cb87319f..., shared as testdata/vectors/tlock_ibe.json; H2-H4 are cited to drand/kyber. - Step 5 makes the SEALED_CONTROL read mandatory, step 15 names ERR_HEADER_BINDING, §21 makes capsule_id 16 CSPRNG bytes a MUST, §76 is made accurate (four dk1.json vectors, the §36 time_only rule, two cases rewritten against the texts that really existed), and editorial fixes in §5, §36, §55.1, §69.1 and §77. §73 lists the three new decisions. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks ago
| `vectors/tlock_ibe.json` | H2 of the tlock IBE: the serialization of an element of GT | §63 step 11 |
Implement capsule format 2 of spec v0.9 The reference moves to the DateKeys Protocol Specification v0.9, approved by its author on 29 September 2026. Encrypt writes capsule format 2 only; Open and Inspect read formats 1 and 2, and a format 1 capsule keeps the verdict v0.8.2 gave it. Format 2 (spec §22, §29.1, §31, §39): - VERSION in the PRELUDE is the capsule format, capsule.Format; any other value is ERR_UNSUPPORTED_VERSION at step 2. - CONTROL_CBOR has the schema version of its format. Version 2 adds key 6, payload_length (8 bytes, big-endian, at most L_MAX = 2^53 - 2^46), and key 7, padding (1 bloque256, 2 reforzado); it is 103 bytes without extensions, whatever L. - The payload is the content padded with zeros to P = rule(L). Step 17 checks the length and the zeros, and Open writes only the first L bytes. - INNER_ACCESS_AGE holds exactly 16 X25519 stanzas: 1 to 16 credentials, and a dummy in each slot left, in a uniformly random order. Writer rules (spec §62.1): EncryptOptions.Length is required and the source must deliver exactly that many bytes; recipients that are not canonical or of low order are rejected (agewrap.CheckX25519Recipient); self-checks of the header, the control, INNER_ACCESS_AGE and PAYLOAD_AGE. The CLI measures its input, takes -padding and reports the format. Test data: seven format 2 fixtures, padding vectors checked against math/big, format 2 CBOR vectors, and the mutation corpus in both formats with the 22 cases of the third list of spec §64, built without randomness by sealing the fixtures again with their known keys and nonces. The format 1 fixtures are kept byte for byte and never regenerated; the differential corpus keeps its 1825 cases and adds a block per format 2 fixture. The spec copy loses its "to be implemented" markers, and the READMEs, CHANGELOG, traceability and testdata/README.md follow v0.9. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
| `vectors/padding.json` | the padding of format 2: P for each content length L, and the length of PAYLOAD_AGE | §29.1 |
Format 3, step 3: the reader Open reads format 3 (spec 29.2 to 29.7, 63 steps 17 and 18): the PRELUDE accepts VERSION 3, and the files go to a Sink. - Sink: Begin with the validated head, Create for each file in the order of the head, and Commit only after every check of step 17; after any failure that follows a successful Begin, Abort, once. A format 3 capsule without a Sink fails right after step 2 with ErrSinkRequired, a caller error with no code, no failed step and no request; a capsule of format 1 or 2 without dst fails there too. - Step 17 in its substeps: the frame and the area, security and its verdicts, which never fail, the head, the files filling CONTENT, the SHA-256 of each file and the padding. A failure of age or a plaintext whose length is not P prevails; otherwise the first substep that fails decides, and a code other than ERR_INTEGRITY is reported only after reading PAYLOAD_AGE to its end. - The reads of BODY grow with the bytes received, never with AREA_LEN, HEAD_LEN or a declared size (spec 57); a test measures it. - A failure of the Sink is the caller's own error with ERR_INTEGRITY, as one of dst is in formats 1 and 2. - Opened gains Head, Verdicts, AreaLen and UnusableHeadExtensions. - Test data: "version changed" sets VERSION 4, and the format 2 list gains "format 2 time_only relabeled format 3", which fails at step 14, as section 64 of spec v0.10 lists: 126 cases, 89 of the spec. The randomly built capsules keep their recorded bytes. - testkit: Build writes format 3 and can edit the padded plaintext; Head3, Body3, DiscardSink and MemorySink build and open BODY. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
| `vectors/mutations.json` | the mutation corpus: the 89 mutations of §64 and further cases | §63, §64 |
Implement capsule format 2 of spec v0.9 The reference moves to the DateKeys Protocol Specification v0.9, approved by its author on 29 September 2026. Encrypt writes capsule format 2 only; Open and Inspect read formats 1 and 2, and a format 1 capsule keeps the verdict v0.8.2 gave it. Format 2 (spec §22, §29.1, §31, §39): - VERSION in the PRELUDE is the capsule format, capsule.Format; any other value is ERR_UNSUPPORTED_VERSION at step 2. - CONTROL_CBOR has the schema version of its format. Version 2 adds key 6, payload_length (8 bytes, big-endian, at most L_MAX = 2^53 - 2^46), and key 7, padding (1 bloque256, 2 reforzado); it is 103 bytes without extensions, whatever L. - The payload is the content padded with zeros to P = rule(L). Step 17 checks the length and the zeros, and Open writes only the first L bytes. - INNER_ACCESS_AGE holds exactly 16 X25519 stanzas: 1 to 16 credentials, and a dummy in each slot left, in a uniformly random order. Writer rules (spec §62.1): EncryptOptions.Length is required and the source must deliver exactly that many bytes; recipients that are not canonical or of low order are rejected (agewrap.CheckX25519Recipient); self-checks of the header, the control, INNER_ACCESS_AGE and PAYLOAD_AGE. The CLI measures its input, takes -padding and reports the format. Test data: seven format 2 fixtures, padding vectors checked against math/big, format 2 CBOR vectors, and the mutation corpus in both formats with the 22 cases of the third list of spec §64, built without randomness by sealing the fixtures again with their known keys and nonces. The format 1 fixtures are kept byte for byte and never regenerated; the differential corpus keeps its 1825 cases and adds a block per format 2 fixture. The spec copy loses its "to be implemented" markers, and the READMEs, CHANGELOG, traceability and testdata/README.md follow v0.9. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
| `vectors/inspect_differential.json` | 4380 mutations of the fixtures with the verdict of steps 1 to 8 | §63 |
| `fixtures/<name>.dkc`, `<name>.json` | official capsules and every intermediate value | §67 |
| `fixtures/<name>.dkk`, `<name>.dkk.json` | official access keys | §68 |
Implement capsule format 2 of spec v0.9 The reference moves to the DateKeys Protocol Specification v0.9, approved by its author on 29 September 2026. Encrypt writes capsule format 2 only; Open and Inspect read formats 1 and 2, and a format 1 capsule keeps the verdict v0.8.2 gave it. Format 2 (spec §22, §29.1, §31, §39): - VERSION in the PRELUDE is the capsule format, capsule.Format; any other value is ERR_UNSUPPORTED_VERSION at step 2. - CONTROL_CBOR has the schema version of its format. Version 2 adds key 6, payload_length (8 bytes, big-endian, at most L_MAX = 2^53 - 2^46), and key 7, padding (1 bloque256, 2 reforzado); it is 103 bytes without extensions, whatever L. - The payload is the content padded with zeros to P = rule(L). Step 17 checks the length and the zeros, and Open writes only the first L bytes. - INNER_ACCESS_AGE holds exactly 16 X25519 stanzas: 1 to 16 credentials, and a dummy in each slot left, in a uniformly random order. Writer rules (spec §62.1): EncryptOptions.Length is required and the source must deliver exactly that many bytes; recipients that are not canonical or of low order are rejected (agewrap.CheckX25519Recipient); self-checks of the header, the control, INNER_ACCESS_AGE and PAYLOAD_AGE. The CLI measures its input, takes -padding and reports the format. Test data: seven format 2 fixtures, padding vectors checked against math/big, format 2 CBOR vectors, and the mutation corpus in both formats with the 22 cases of the third list of spec §64, built without randomness by sealing the fixtures again with their known keys and nonces. The format 1 fixtures are kept byte for byte and never regenerated; the differential corpus keeps its 1825 cases and adds a block per format 2 fixture. The spec copy loses its "to be implemented" markers, and the READMEs, CHANGELOG, traceability and testdata/README.md follow v0.9. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
| `fixtures/<name>.plaintext` | the content of each capsule, what the reader delivers: in format 2, without its padding | §67 |
| `fixtures/<name>.inspect.json` | the exact output of `datekeys inspect -json` for each `.dkc` | §63 |
Implement capsule format 2 of spec v0.9 The reference moves to the DateKeys Protocol Specification v0.9, approved by its author on 29 September 2026. Encrypt writes capsule format 2 only; Open and Inspect read formats 1 and 2, and a format 1 capsule keeps the verdict v0.8.2 gave it. Format 2 (spec §22, §29.1, §31, §39): - VERSION in the PRELUDE is the capsule format, capsule.Format; any other value is ERR_UNSUPPORTED_VERSION at step 2. - CONTROL_CBOR has the schema version of its format. Version 2 adds key 6, payload_length (8 bytes, big-endian, at most L_MAX = 2^53 - 2^46), and key 7, padding (1 bloque256, 2 reforzado); it is 103 bytes without extensions, whatever L. - The payload is the content padded with zeros to P = rule(L). Step 17 checks the length and the zeros, and Open writes only the first L bytes. - INNER_ACCESS_AGE holds exactly 16 X25519 stanzas: 1 to 16 credentials, and a dummy in each slot left, in a uniformly random order. Writer rules (spec §62.1): EncryptOptions.Length is required and the source must deliver exactly that many bytes; recipients that are not canonical or of low order are rejected (agewrap.CheckX25519Recipient); self-checks of the header, the control, INNER_ACCESS_AGE and PAYLOAD_AGE. The CLI measures its input, takes -padding and reports the format. Test data: seven format 2 fixtures, padding vectors checked against math/big, format 2 CBOR vectors, and the mutation corpus in both formats with the 22 cases of the third list of spec §64, built without randomness by sealing the fixtures again with their known keys and nonces. The format 1 fixtures are kept byte for byte and never regenerated; the differential corpus keeps its 1825 cases and adds a block per format 2 fixture. The spec copy loses its "to be implemented" markers, and the READMEs, CHANGELOG, traceability and testdata/README.md follow v0.9. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
There are twelve official capsules. Five are in format 1, the fixtures of
v0.8.2, kept for compatibility: `time_only`, `time_only_extensions`,
`time_and_key_portable`, `time_and_key_recipients` and `empty_payload`. Seven
are in format 2:
| Fixture | Policy | Credentials | L | Padding code | P |
|---|---|---|---|---|---|
| `format2_time_only` | `time_only` | — | 78000 | 2, reforzado | 79872 |
| `format2_time_only_bloque256` | `time_only` | — | 78000, the same content | 1, bloque256 | 78080 |
| `format2_empty_payload` | `time_only` | — | 0 | 2 | 256 |
| `format2_time_only_extensions` | `time_only` | — | 34 | 2 | 256 |
| `format2_time_and_key_portable` | `time_and_key` | a portable `.dkk`; 15 dummies | 46 | 2 | 256 |
| `format2_time_and_key_recipients` | `time_and_key` | a portable `.dkk` and three X25519 identities; 12 dummies | 41 | 2 | 256 |
| `format2_time_and_key_sixteen` | `time_and_key` | sixteen X25519 identities; no dummy | 41 | 2 | 256 |
The first two have a P above 64 KiB, so their PAYLOAD_AGE has two STREAM
chunks. `format2_time_only_extensions` carries a noncritical PUBLIC_HEADER
extension and a noncritical CONTROL_CBOR extension. The release that opens each
capsule, a published Quicknet signature, is in its `<name>.json`, so they all
decrypt offline.
## Edited files
`mutations.json` and `inspect_differential.json` give each mutated `.dkc` as
edits of a base file, not as its full bytes:
```json
Format 3, step 3: the reader Open reads format 3 (spec 29.2 to 29.7, 63 steps 17 and 18): the PRELUDE accepts VERSION 3, and the files go to a Sink. - Sink: Begin with the validated head, Create for each file in the order of the head, and Commit only after every check of step 17; after any failure that follows a successful Begin, Abort, once. A format 3 capsule without a Sink fails right after step 2 with ErrSinkRequired, a caller error with no code, no failed step and no request; a capsule of format 1 or 2 without dst fails there too. - Step 17 in its substeps: the frame and the area, security and its verdicts, which never fail, the head, the files filling CONTENT, the SHA-256 of each file and the padding. A failure of age or a plaintext whose length is not P prevails; otherwise the first substep that fails decides, and a code other than ERR_INTEGRITY is reported only after reading PAYLOAD_AGE to its end. - The reads of BODY grow with the bytes received, never with AREA_LEN, HEAD_LEN or a declared size (spec 57); a test measures it. - A failure of the Sink is the caller's own error with ERR_INTEGRITY, as one of dst is in formats 1 and 2. - Opened gains Head, Verdicts, AreaLen and UnusableHeadExtensions. - Test data: "version changed" sets VERSION 4, and the format 2 list gains "format 2 time_only relabeled format 3", which fails at step 14, as section 64 of spec v0.10 lists: 126 cases, 89 of the spec. The randomly built capsules keep their recorded bytes. - testkit: Build writes format 3 and can edit the padded plaintext; Head3, Body3, DiscardSink and MemorySink build and open BODY. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
{ "base": "time_only.dkc", "edits": [[4, 1, "04"]] }
```
- `base` is a file of `testdata/fixtures`. In `mutations.json` it may be absent:
the base is then the empty file, and the single edit holds the whole capsule.
- An edit is `[at, delete, insert]`: the `delete` bytes at offset `at` of the
base are replaced by the bytes of the hex string `insert`.
- The edits of one file refer to offsets of the unmodified base, are sorted by
`at` and do not overlap. The result is therefore built in one pass: copy the
base up to `at`, append `insert`, skip `delete` bytes of the base, go on with
the next edit, and copy the rest of the base.
- `[0, 78799, ""]` on `time_only.dkc` is the empty file; `"edits": []` is the
base unchanged.
## `vectors/cbor.json`
```json
{
Implement capsule format 2 of spec v0.9 The reference moves to the DateKeys Protocol Specification v0.9, approved by its author on 29 September 2026. Encrypt writes capsule format 2 only; Open and Inspect read formats 1 and 2, and a format 1 capsule keeps the verdict v0.8.2 gave it. Format 2 (spec §22, §29.1, §31, §39): - VERSION in the PRELUDE is the capsule format, capsule.Format; any other value is ERR_UNSUPPORTED_VERSION at step 2. - CONTROL_CBOR has the schema version of its format. Version 2 adds key 6, payload_length (8 bytes, big-endian, at most L_MAX = 2^53 - 2^46), and key 7, padding (1 bloque256, 2 reforzado); it is 103 bytes without extensions, whatever L. - The payload is the content padded with zeros to P = rule(L). Step 17 checks the length and the zeros, and Open writes only the first L bytes. - INNER_ACCESS_AGE holds exactly 16 X25519 stanzas: 1 to 16 credentials, and a dummy in each slot left, in a uniformly random order. Writer rules (spec §62.1): EncryptOptions.Length is required and the source must deliver exactly that many bytes; recipients that are not canonical or of low order are rejected (agewrap.CheckX25519Recipient); self-checks of the header, the control, INNER_ACCESS_AGE and PAYLOAD_AGE. The CLI measures its input, takes -padding and reports the format. Test data: seven format 2 fixtures, padding vectors checked against math/big, format 2 CBOR vectors, and the mutation corpus in both formats with the 22 cases of the third list of spec §64, built without randomness by sealing the fixtures again with their known keys and nonces. The format 1 fixtures are kept byte for byte and never regenerated; the differential corpus keeps its 1825 cases and adds a block per format 2 fixture. The spec copy loses its "to be implemented" markers, and the READMEs, CHANGELOG, traceability and testdata/README.md follow v0.9. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
"spec": "0.9",
"walk": { "max_depth": 3, "max_len": 64 },
"accept": [ { "name": "uint 2^53 eight bytes", "hex": "1b0020000000000000", "value": "9007199254740992" } ],
"reject": [ { "name": "tag", "hex": "c101", "error": "ERR_NON_CANONICAL_CBOR" } ],
"schemas": [ { "block": "extension", "schema": "public_header", "name": "65 extensions", "hex": "a600…", "result": "ERR_NON_CANONICAL_CBOR" } ]
}
```
### Generic vectors: `accept` and `reject`
Each `hex` is checked as exactly one data item of the CBOR profile of spec §58:
major types 0, 2, 3, 4 and 5 only; integers and lengths in their shortest form;
definite lengths; map keys that are unsigned integers in strictly ascending
order; valid UTF-8 text; nothing after the item. `accept` holds the inputs that
pass, `reject` those that fail, all with `ERR_NON_CANONICAL_CBOR`.
The `walk` limits apply as well, as in the reference `codec.Walk`: containers
nest at most `max_depth` deep (a scalar has depth 0, `81818100` has depth 3),
and every byte string and text string has at most `max_len` bytes, every array
at most `max_len` items and every map at most `max_len` entries. The vectors
named "above max_len" or "above max_depth" fail on these limits only.
`value` is present for an accepted unsigned integer: a JSON number up to
2^53 − 1, and a decimal string above, so that no reader loses precision.
Among them: shortest-form boundaries, `a200010101` (the two-key map
`{0: 1, 1: 1}`), a text with a leading BOM, keys out of order or repeated,
non-integer keys, indefinite lengths, tags, floats, simple values, negative
integers, truncation, lengths beyond the input, trailing bytes, overlong UTF-8
and surrogates.
### Schema vectors: `schemas`
Each vector is one encoded object:
- `schema` names the object and the decoder to run on `hex`:
- `provider_profile`: a Provider Profile (§11), decoded and then validated
Spec v0.8.2 refinements: error precedence, trust model, strict order Approved refinements, each recorded with its reproducible case in the §76 v0.8.2 subsection: - §69.1: layered error model with normative precedence (frame, type tag and version, CBOR profile and CDDL, then fields with their own code in ascending key order; across steps the §63 order decides), with a scope paragraph for the optional steps 5, 6 and 8. - §55.1: normative trust table per section (who can write it, from which step it is bound, what it never proves); §72: security-relevant claims go in CONTROL_CBOR or under a signature, .dkk data is advisory. - §31/§54: extension arrays in strictly ascending unsigned byte order of extension_id (one rule for order and uniqueness). - Gaps a second implementation needed: §28.1 malformed age headers, §15/§19 latest unlock time and dk1_ reading rules, §22/§23/§57 length lower bounds, §63 step 8 tlock argument comparison and step 9 order, §12.1 profile validation with the drand chain-hash formula, §74 table of implementation limits. Reference alignment: .dkk errors only at step 9.a (new OpenOptions.AccessKeyFile, used by the CLI), CR/LF in dk1_ is ERR_DATEKEY_INVALID, BODY_LEN 0 is ERR_INTEGRITY, nil identities are not credentials, and AccessIdentity tries every identity on every stanza so its verdict does not depend on their order. dk1.json gains three vectors; every other testdata file is byte-identical. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks ago
as a profile to pin, by rules 1 to 3 of spec §12.1 in their order: the
CDDL with the `period` limit of the reference, the field rules
(`ERR_UNKNOWN_PROFILE`) and the chain-hash self-check
(`ERR_PROFILE_MISMATCH`), whose formula §12.1 gives. No `profile_hash` is
expected (rule 4). A vector that changes a hashed key recomputes
`chain_hash`, unless its name says that the chain hash no longer matches.
- `public_header`: PUBLIC_HEADER (§24). No profile registry is consulted and
no extension is known: a header naming an unpinned profile is valid here
(`ERR_UNKNOWN_PROFILE` comes from the registry at step 4), and critical
extensions are not checked here.
Implement capsule format 2 of spec v0.9 The reference moves to the DateKeys Protocol Specification v0.9, approved by its author on 29 September 2026. Encrypt writes capsule format 2 only; Open and Inspect read formats 1 and 2, and a format 1 capsule keeps the verdict v0.8.2 gave it. Format 2 (spec §22, §29.1, §31, §39): - VERSION in the PRELUDE is the capsule format, capsule.Format; any other value is ERR_UNSUPPORTED_VERSION at step 2. - CONTROL_CBOR has the schema version of its format. Version 2 adds key 6, payload_length (8 bytes, big-endian, at most L_MAX = 2^53 - 2^46), and key 7, padding (1 bloque256, 2 reforzado); it is 103 bytes without extensions, whatever L. - The payload is the content padded with zeros to P = rule(L). Step 17 checks the length and the zeros, and Open writes only the first L bytes. - INNER_ACCESS_AGE holds exactly 16 X25519 stanzas: 1 to 16 credentials, and a dummy in each slot left, in a uniformly random order. Writer rules (spec §62.1): EncryptOptions.Length is required and the source must deliver exactly that many bytes; recipients that are not canonical or of low order are rejected (agewrap.CheckX25519Recipient); self-checks of the header, the control, INNER_ACCESS_AGE and PAYLOAD_AGE. The CLI measures its input, takes -padding and reports the format. Test data: seven format 2 fixtures, padding vectors checked against math/big, format 2 CBOR vectors, and the mutation corpus in both formats with the 22 cases of the third list of spec §64, built without randomness by sealing the fixtures again with their known keys and nonces. The format 1 fixtures are kept byte for byte and never regenerated; the differential corpus keeps its 1825 cases and adds a block per format 2 fixture. The spec copy loses its "to be implemented" markers, and the READMEs, CHANGELOG, traceability and testdata/README.md follow v0.9. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
- `control_cbor`: CONTROL_CBOR (§31), decoded for the capsule format given
by `format`, 1 when the field is absent: the schema version must equal it
(another is `ERR_UNSUPPORTED_VERSION`, layer 2), and keys 6,
`payload_length`, and 7, `padding`, are required in format 2 and not
defined in format 1. The names of the format 2 vectors start with
"format 2: ".
- `dkk_body`: BODY_CBOR of a `.dkk` (§41), without the 12-byte DKK1
prelude.
- `block` names the schema the vector exercises: the same as `schema`, or
`verification_metadata` (the `.dkk` body's key 6 varies) or `extension` (the
PUBLIC_HEADER's key 6, `noncritical_extensions`, varies).
- `result` is `ok` or the error code.
Spec v0.8.2 refinements: error precedence, trust model, strict order Approved refinements, each recorded with its reproducible case in the §76 v0.8.2 subsection: - §69.1: layered error model with normative precedence (frame, type tag and version, CBOR profile and CDDL, then fields with their own code in ascending key order; across steps the §63 order decides), with a scope paragraph for the optional steps 5, 6 and 8. - §55.1: normative trust table per section (who can write it, from which step it is bound, what it never proves); §72: security-relevant claims go in CONTROL_CBOR or under a signature, .dkk data is advisory. - §31/§54: extension arrays in strictly ascending unsigned byte order of extension_id (one rule for order and uniqueness). - Gaps a second implementation needed: §28.1 malformed age headers, §15/§19 latest unlock time and dk1_ reading rules, §22/§23/§57 length lower bounds, §63 step 8 tlock argument comparison and step 9 order, §12.1 profile validation with the drand chain-hash formula, §74 table of implementation limits. Reference alignment: .dkk errors only at step 9.a (new OpenOptions.AccessKeyFile, used by the CLI), CR/LF in dk1_ is ERR_DATEKEY_INVALID, BODY_LEN 0 is ERR_INTEGRITY, nil identities are not credentials, and AccessIdentity tries every identity on every stanza so its verdict does not depend on their order. dk1.json gains three vectors; every other testdata file is byte-identical. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks ago
When bytes break several rules, the code is the one of the first failing
layer of spec §69.1: the type tag and the schema version, read first from keys
0 and 1 (layer 2), then the CDDL, the implementation limits included, and the
re-encoding (layer 3), and only then the fields with codes of their own in
ascending key order (layer 4). The objects of these vectors have no frame, so
layer 1 does not apply, and their critical extensions are not checked (see
above).
Each block has a minimal valid object, an unknown key, a missing required key, a
wrong type and values out of size or range. The `extension` block has, besides:
data `40` (empty) and `5801xx` (length not in its shortest form), data of every
other type (text, `null`, integer, map, array, tag, indefinite length), 64 and
65 extensions, an `extension_id` starting with a BOM, the pair U+FF61 and
U+10000 in UTF-8 byte order (valid) and in UTF-16 order (invalid), and
`extension_version` 2^32 − 1 (valid), 2^32 and 2^53 (invalid).
Implement capsule format 2 of spec v0.9 The reference moves to the DateKeys Protocol Specification v0.9, approved by its author on 29 September 2026. Encrypt writes capsule format 2 only; Open and Inspect read formats 1 and 2, and a format 1 capsule keeps the verdict v0.8.2 gave it. Format 2 (spec §22, §29.1, §31, §39): - VERSION in the PRELUDE is the capsule format, capsule.Format; any other value is ERR_UNSUPPORTED_VERSION at step 2. - CONTROL_CBOR has the schema version of its format. Version 2 adds key 6, payload_length (8 bytes, big-endian, at most L_MAX = 2^53 - 2^46), and key 7, padding (1 bloque256, 2 reforzado); it is 103 bytes without extensions, whatever L. - The payload is the content padded with zeros to P = rule(L). Step 17 checks the length and the zeros, and Open writes only the first L bytes. - INNER_ACCESS_AGE holds exactly 16 X25519 stanzas: 1 to 16 credentials, and a dummy in each slot left, in a uniformly random order. Writer rules (spec §62.1): EncryptOptions.Length is required and the source must deliver exactly that many bytes; recipients that are not canonical or of low order are rejected (agewrap.CheckX25519Recipient); self-checks of the header, the control, INNER_ACCESS_AGE and PAYLOAD_AGE. The CLI measures its input, takes -padding and reports the format. Test data: seven format 2 fixtures, padding vectors checked against math/big, format 2 CBOR vectors, and the mutation corpus in both formats with the 22 cases of the third list of spec §64, built without randomness by sealing the fixtures again with their known keys and nonces. The format 1 fixtures are kept byte for byte and never regenerated; the differential corpus keeps its 1825 cases and adds a block per format 2 fixture. The spec copy loses its "to be implemented" markers, and the READMEs, CHANGELOG, traceability and testdata/README.md follow v0.9. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
The `control_cbor` block has, besides, a key 6 in format 1, where it is not
defined, and in format 2: the minimal control of 103 bytes (L = 0),
`payload_length` 78000, 2^32 + 1 and L_MAX = 2^53 − 2^46 (valid), L_MAX + 1,
2^53 and 2^64 − 1 (invalid), a `payload_length` of 7 or 9 bytes, as an
unsigned integer or with a length not in its shortest form, `padding` 1
(valid), 0, 3, 257 and as a byte string, a missing key 6 or 7, keys 6 and 7
out of order, an unknown key 8, and schema versions 1 and 3
(`ERR_UNSUPPORTED_VERSION`).
#### Implementation limits
Spec v0.8.2 refinements: error precedence, trust model, strict order Approved refinements, each recorded with its reproducible case in the §76 v0.8.2 subsection: - §69.1: layered error model with normative precedence (frame, type tag and version, CBOR profile and CDDL, then fields with their own code in ascending key order; across steps the §63 order decides), with a scope paragraph for the optional steps 5, 6 and 8. - §55.1: normative trust table per section (who can write it, from which step it is bound, what it never proves); §72: security-relevant claims go in CONTROL_CBOR or under a signature, .dkk data is advisory. - §31/§54: extension arrays in strictly ascending unsigned byte order of extension_id (one rule for order and uniqueness). - Gaps a second implementation needed: §28.1 malformed age headers, §15/§19 latest unlock time and dk1_ reading rules, §22/§23/§57 length lower bounds, §63 step 8 tlock argument comparison and step 9 order, §12.1 profile validation with the drand chain-hash formula, §74 table of implementation limits. Reference alignment: .dkk errors only at step 9.a (new OpenOptions.AccessKeyFile, used by the CLI), CR/LF in dk1_ is ERR_DATEKEY_INVALID, BODY_LEN 0 is ERR_INTEGRITY, nil identities are not credentials, and AccessIdentity tries every identity on every stanza so its verdict does not depend on their order. dk1.json gains three vectors; every other testdata file is byte-identical. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks ago
The vectors apply the implementation limits of the reference that spec §74
lists: the maximum `extension_id` length, the Provider Profile `period` of one
day, the maximum name lengths and the `public_key` length. The vectors named
"the implementation limit" sit at a limit and are valid; those named "above
the implementation limit" go one past it and are otherwise valid, so that an
implementation without the limit accepts them. The name alphabets, the
`genesis_time` range and the drand rules of the `provider_profile`
validation are not limits: they are rules of spec §12.1. Neither is the
minimum of one byte of `extension_id` (spec §31).
Spec v0.8.2: corrections from the formal review A formal review of the whole v0.8.2 text found it approvable after these corrections, recorded in §76 ("Correcciones de la revisión formal"): - §27 no longer calls header_binding the authenticity of PUBLIC_HEADER: it binds the header to the opened control, never authorship or date (§55.1); the age MAC only protects against whoever lacks the file key. - §63 steps 9 and 10: a network source (relay, Release API, cache) MUST verify every response and gives ERR_RELEASE_UNAVAILABLE at step 9 when none verifies; the step-10 codes are for a directly supplied release. The reference already behaved so; TestReleaseFromANetworkSource pins both paths. - §54 and §72: registrations declare the objects and arrays where an extension may appear, and a known extension out of place counts as unknown there. The reference gains the optional extension.Placement interface, used at steps 4, 9.a and 14. - §63 step 11 fixes the GT serialization hashed by H2 (kilic/kyber order) with the frozen vector H2(e(G1, G2))[:16] = cb87319f..., shared as testdata/vectors/tlock_ibe.json; H2-H4 are cited to drand/kyber. - Step 5 makes the SEALED_CONTROL read mandatory, step 15 names ERR_HEADER_BINDING, §21 makes capsule_id 16 CSPRNG bytes a MUST, §76 is made accurate (four dk1.json vectors, the §36 time_only rule, two cases rewritten against the texts that really existed), and editorial fixes in §5, §36, §55.1, §69.1 and §77. §73 lists the three new decisions. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks ago
## `vectors/tlock_ibe.json`
H2, the hash of an element of GT in the IBE-CCA with which tlock wraps
FK_TIME: spec §63 step 11, and the paragraph "Serialización de GT en H2" after
the flow.
```json
{
Implement capsule format 2 of spec v0.9 The reference moves to the DateKeys Protocol Specification v0.9, approved by its author on 29 September 2026. Encrypt writes capsule format 2 only; Open and Inspect read formats 1 and 2, and a format 1 capsule keeps the verdict v0.8.2 gave it. Format 2 (spec §22, §29.1, §31, §39): - VERSION in the PRELUDE is the capsule format, capsule.Format; any other value is ERR_UNSUPPORTED_VERSION at step 2. - CONTROL_CBOR has the schema version of its format. Version 2 adds key 6, payload_length (8 bytes, big-endian, at most L_MAX = 2^53 - 2^46), and key 7, padding (1 bloque256, 2 reforzado); it is 103 bytes without extensions, whatever L. - The payload is the content padded with zeros to P = rule(L). Step 17 checks the length and the zeros, and Open writes only the first L bytes. - INNER_ACCESS_AGE holds exactly 16 X25519 stanzas: 1 to 16 credentials, and a dummy in each slot left, in a uniformly random order. Writer rules (spec §62.1): EncryptOptions.Length is required and the source must deliver exactly that many bytes; recipients that are not canonical or of low order are rejected (agewrap.CheckX25519Recipient); self-checks of the header, the control, INNER_ACCESS_AGE and PAYLOAD_AGE. The CLI measures its input, takes -padding and reports the format. Test data: seven format 2 fixtures, padding vectors checked against math/big, format 2 CBOR vectors, and the mutation corpus in both formats with the 22 cases of the third list of spec §64, built without randomness by sealing the fixtures again with their known keys and nonces. The format 1 fixtures are kept byte for byte and never regenerated; the differential corpus keeps its 1825 cases and adds a block per format 2 fixture. The spec copy loses its "to be implemented" markers, and the READMEs, CHANGELOG, traceability and testdata/README.md follow v0.9. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
"spec": "0.9",
Spec v0.8.2: corrections from the formal review A formal review of the whole v0.8.2 text found it approvable after these corrections, recorded in §76 ("Correcciones de la revisión formal"): - §27 no longer calls header_binding the authenticity of PUBLIC_HEADER: it binds the header to the opened control, never authorship or date (§55.1); the age MAC only protects against whoever lacks the file key. - §63 steps 9 and 10: a network source (relay, Release API, cache) MUST verify every response and gives ERR_RELEASE_UNAVAILABLE at step 9 when none verifies; the step-10 codes are for a directly supplied release. The reference already behaved so; TestReleaseFromANetworkSource pins both paths. - §54 and §72: registrations declare the objects and arrays where an extension may appear, and a known extension out of place counts as unknown there. The reference gains the optional extension.Placement interface, used at steps 4, 9.a and 14. - §63 step 11 fixes the GT serialization hashed by H2 (kilic/kyber order) with the frozen vector H2(e(G1, G2))[:16] = cb87319f..., shared as testdata/vectors/tlock_ibe.json; H2-H4 are cited to drand/kyber. - Step 5 makes the SEALED_CONTROL read mandatory, step 15 names ERR_HEADER_BINDING, §21 makes capsule_id 16 CSPRNG bytes a MUST, §76 is made accurate (four dk1.json vectors, the §36 time_only rule, two cases rewritten against the texts that really existed), and editorial fixes in §5, §36, §55.1, §69.1 and §77. §73 lists the three new decisions. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks ago
"description": "…",
"vectors": [
{ "name": "H2(e(G1, G2)), the generators of G1 and G2", "g1": "97f1…", "g2": "93e0…", "gt": "0f41…", "h2": "cb87319f24560b5231579a09ad79f12e" }
]
}
```
- `g1`, `g2`: a point of G1 and one of G2 in the canonical compressed
encoding of spec §12.2, 48 and 96 bytes; the vector uses the generators.
- `gt`: e(g1, g2), the pairing of G1 × G2 of step 11, in 576 bytes. Over the
tower Fp2 = Fp[u]/(u² + 1), Fp6 = Fp2[v]/(v³ − (u + 1)) and
Fp12 = Fp6[w]/(w² − v), every element is written with its coefficients from
the highest degree to the lowest: c1 then c0 for Fp12, c2, c1 and c0 for
Fp6, c1 then c0 for Fp2, and each element of Fp in 48 bytes big-endian. It
is the order of kilic/bls12-381, which drand/kyber and tlock use.
- `h2`: SHA-256 of the ASCII bytes `IBE-H2` followed by `gt`, truncated to
its first 16 bytes, the length of V.
A reader checks that its pairing of `g1` and `g2` serializes to `gt` and that
its H2 of those bytes is `h2`; the vector fixes the pairing and the
serialization at once. The same 576 bytes with the twelve coordinates of Fp in
reverse order, c0 first at every level as `Fp12.toBytes` of `@noble/curves`
writes them, give `0118eea9d5971745f71e3c94926f1717` and another FK_TIME.
Implement capsule format 2 of spec v0.9 The reference moves to the DateKeys Protocol Specification v0.9, approved by its author on 29 September 2026. Encrypt writes capsule format 2 only; Open and Inspect read formats 1 and 2, and a format 1 capsule keeps the verdict v0.8.2 gave it. Format 2 (spec §22, §29.1, §31, §39): - VERSION in the PRELUDE is the capsule format, capsule.Format; any other value is ERR_UNSUPPORTED_VERSION at step 2. - CONTROL_CBOR has the schema version of its format. Version 2 adds key 6, payload_length (8 bytes, big-endian, at most L_MAX = 2^53 - 2^46), and key 7, padding (1 bloque256, 2 reforzado); it is 103 bytes without extensions, whatever L. - The payload is the content padded with zeros to P = rule(L). Step 17 checks the length and the zeros, and Open writes only the first L bytes. - INNER_ACCESS_AGE holds exactly 16 X25519 stanzas: 1 to 16 credentials, and a dummy in each slot left, in a uniformly random order. Writer rules (spec §62.1): EncryptOptions.Length is required and the source must deliver exactly that many bytes; recipients that are not canonical or of low order are rejected (agewrap.CheckX25519Recipient); self-checks of the header, the control, INNER_ACCESS_AGE and PAYLOAD_AGE. The CLI measures its input, takes -padding and reports the format. Test data: seven format 2 fixtures, padding vectors checked against math/big, format 2 CBOR vectors, and the mutation corpus in both formats with the 22 cases of the third list of spec §64, built without randomness by sealing the fixtures again with their known keys and nonces. The format 1 fixtures are kept byte for byte and never regenerated; the differential corpus keeps its 1825 cases and adds a block per format 2 fixture. The spec copy loses its "to be implemented" markers, and the READMEs, CHANGELOG, traceability and testdata/README.md follow v0.9. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
## `vectors/padding.json`
The padding of the payload of a format 2 capsule, spec §29.1: for each content
length L, the length P of the plaintext of PAYLOAD_AGE with each code, and the
length of PAYLOAD_AGE.
```json
{
"spec": "0.9",
"l_max": 8936830510563328,
"vectors": [
{ "l": 78000, "bloque256": 78080, "reforzado": 79872, "payload_age_bloque256": 78296, "payload_age_reforzado": 80088, "e": 16, "s": 5, "last_bits": 11 }
],
"rejected": [8936830510563329, 9007199254740992]
}
```
- `l`: L, the length of the content.
- `bloque256`: P with code 1, 256 · ⌈L / 256⌉, at least 256. `reforzado`: P
with code 2, the larger of that and Padmé.
- `payload_age_bloque256`, `payload_age_reforzado`: the length of PAYLOAD_AGE
for each P, 184 + P + 16 · max(1, ⌈P / 65536⌉): an age header of one X25519
stanza (168 bytes), the 16-byte nonce, and the P bytes in STREAM chunks of
64 KiB with a 16-byte tag each.
- `e`, `s`, `last_bits`: the intermediate values of Padmé, E = bitlen(L) − 1,
S = bitlen(E) and E − S; only for L > 256, and informative.
- `l_max`: L_MAX = 2^53 − 2^46, the largest L.
- `rejected`: lengths above L_MAX, which a writer rejects and which, as key 6
of a control, are `ERR_NON_CANONICAL_CBOR`.
Every number is below 2^53, so a JSON number holds it exactly. The lengths
include both sides of the boundaries where 32-bit arithmetic breaks
(2^31 − 2^25, 2^32 − 2^26, 2^32) and 2^49 − 1, the first L for which a
floating-point logarithm gives an E one too large (P does not change).
## `vectors/mutations.json`
The mutation corpus of spec §64, as frozen data. Each case is a `.dkc`, what
the reader is given to open it, and the exact error and step at which the full
reading flow (`capsule.Open`, §63) must fail.
```json
{
"name": "version changed",
"spec": true,
Format 3, step 3: the reader Open reads format 3 (spec 29.2 to 29.7, 63 steps 17 and 18): the PRELUDE accepts VERSION 3, and the files go to a Sink. - Sink: Begin with the validated head, Create for each file in the order of the head, and Commit only after every check of step 17; after any failure that follows a successful Begin, Abort, once. A format 3 capsule without a Sink fails right after step 2 with ErrSinkRequired, a caller error with no code, no failed step and no request; a capsule of format 1 or 2 without dst fails there too. - Step 17 in its substeps: the frame and the area, security and its verdicts, which never fail, the head, the files filling CONTENT, the SHA-256 of each file and the padding. A failure of age or a plaintext whose length is not P prevails; otherwise the first substep that fails decides, and a code other than ERR_INTEGRITY is reported only after reading PAYLOAD_AGE to its end. - The reads of BODY grow with the bytes received, never with AREA_LEN, HEAD_LEN or a declared size (spec 57); a test measures it. - A failure of the Sink is the caller's own error with ERR_INTEGRITY, as one of dst is in formats 1 and 2. - Opened gains Head, Verdicts, AreaLen and UnusableHeadExtensions. - Test data: "version changed" sets VERSION 4, and the format 2 list gains "format 2 time_only relabeled format 3", which fails at step 14, as section 64 of spec v0.10 lists: 126 cases, 89 of the spec. The randomly built capsules keep their recorded bytes. - testkit: Build writes format 3 and can edit the padded plaintext; Head3, Body3, DiscardSink and MemorySink build and open BODY. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
"dkc": { "base": "time_only.dkc", "edits": [[4, 1, "04"]] },
"release": { "round": 1000, "signature": "b446…" },
"now": "2023-08-23T15:59:24Z",
"registry": "default",
"network": false,
"frozen": false,
"error": "ERR_UNSUPPORTED_VERSION",
"step": 2
}
```
- `name`: unique, stable.
Format 3, step 3: the reader Open reads format 3 (spec 29.2 to 29.7, 63 steps 17 and 18): the PRELUDE accepts VERSION 3, and the files go to a Sink. - Sink: Begin with the validated head, Create for each file in the order of the head, and Commit only after every check of step 17; after any failure that follows a successful Begin, Abort, once. A format 3 capsule without a Sink fails right after step 2 with ErrSinkRequired, a caller error with no code, no failed step and no request; a capsule of format 1 or 2 without dst fails there too. - Step 17 in its substeps: the frame and the area, security and its verdicts, which never fail, the head, the files filling CONTENT, the SHA-256 of each file and the padding. A failure of age or a plaintext whose length is not P prevails; otherwise the first substep that fails decides, and a code other than ERR_INTEGRITY is reported only after reading PAYLOAD_AGE to its end. - The reads of BODY grow with the bytes received, never with AREA_LEN, HEAD_LEN or a declared size (spec 57); a test measures it. - A failure of the Sink is the caller's own error with ERR_INTEGRITY, as one of dst is in formats 1 and 2. - Opened gains Head, Verdicts, AreaLen and UnusableHeadExtensions. - Test data: "version changed" sets VERSION 4, and the format 2 list gains "format 2 time_only relabeled format 3", which fails at step 14, as section 64 of spec v0.10 lists: 126 cases, 89 of the spec. The randomly built capsules keep their recorded bytes. - testkit: Build writes format 3 and can edit the padded plaintext; Head3, Body3, DiscardSink and MemorySink build and open BODY. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
- `spec`: true for the 89 mutations listed in spec §64, false for the further
Implement capsule format 2 of spec v0.9 The reference moves to the DateKeys Protocol Specification v0.9, approved by its author on 29 September 2026. Encrypt writes capsule format 2 only; Open and Inspect read formats 1 and 2, and a format 1 capsule keeps the verdict v0.8.2 gave it. Format 2 (spec §22, §29.1, §31, §39): - VERSION in the PRELUDE is the capsule format, capsule.Format; any other value is ERR_UNSUPPORTED_VERSION at step 2. - CONTROL_CBOR has the schema version of its format. Version 2 adds key 6, payload_length (8 bytes, big-endian, at most L_MAX = 2^53 - 2^46), and key 7, padding (1 bloque256, 2 reforzado); it is 103 bytes without extensions, whatever L. - The payload is the content padded with zeros to P = rule(L). Step 17 checks the length and the zeros, and Open writes only the first L bytes. - INNER_ACCESS_AGE holds exactly 16 X25519 stanzas: 1 to 16 credentials, and a dummy in each slot left, in a uniformly random order. Writer rules (spec §62.1): EncryptOptions.Length is required and the source must deliver exactly that many bytes; recipients that are not canonical or of low order are rejected (agewrap.CheckX25519Recipient); self-checks of the header, the control, INNER_ACCESS_AGE and PAYLOAD_AGE. The CLI measures its input, takes -padding and reports the format. Test data: seven format 2 fixtures, padding vectors checked against math/big, format 2 CBOR vectors, and the mutation corpus in both formats with the 22 cases of the third list of spec §64, built without randomness by sealing the fixtures again with their known keys and nonces. The format 1 fixtures are kept byte for byte and never regenerated; the differential corpus keeps its 1825 cases and adds a block per format 2 fixture. The spec copy loses its "to be implemented" markers, and the READMEs, CHANGELOG, traceability and testdata/README.md follow v0.9. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
cases of the reference. The cases come in this order: the 33 mutations of
the first two lists of §64 on the format 1 fixtures (cases 1 to 33), 32
further cases (34 to 65), the same 33 mutations on the format 2 fixtures,
Format 3, step 3: the reader Open reads format 3 (spec 29.2 to 29.7, 63 steps 17 and 18): the PRELUDE accepts VERSION 3, and the files go to a Sink. - Sink: Begin with the validated head, Create for each file in the order of the head, and Commit only after every check of step 17; after any failure that follows a successful Begin, Abort, once. A format 3 capsule without a Sink fails right after step 2 with ErrSinkRequired, a caller error with no code, no failed step and no request; a capsule of format 1 or 2 without dst fails there too. - Step 17 in its substeps: the frame and the area, security and its verdicts, which never fail, the head, the files filling CONTENT, the SHA-256 of each file and the padding. A failure of age or a plaintext whose length is not P prevails; otherwise the first substep that fails decides, and a code other than ERR_INTEGRITY is reported only after reading PAYLOAD_AGE to its end. - The reads of BODY grow with the bytes received, never with AREA_LEN, HEAD_LEN or a declared size (spec 57); a test measures it. - A failure of the Sink is the caller's own error with ERR_INTEGRITY, as one of dst is in formats 1 and 2. - Opened gains Head, Verdicts, AreaLen and UnusableHeadExtensions. - Test data: "version changed" sets VERSION 4, and the format 2 list gains "format 2 time_only relabeled format 3", which fails at step 14, as section 64 of spec v0.10 lists: 126 cases, 89 of the spec. The randomly built capsules keep their recorded bytes. - testkit: Build writes format 3 and can edit the padded plaintext; Head3, Body3, DiscardSink and MemorySink build and open BODY. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
named "format 2: …" (66 to 98), the 23 of the third list (99 to 121), and 5
further cases (122 to 126).
- `dkc`: the capsule, as edits of a fixture (see above). The reader gets it as a
Spec v0.8.2 refinements: error precedence, trust model, strict order Approved refinements, each recorded with its reproducible case in the §76 v0.8.2 subsection: - §69.1: layered error model with normative precedence (frame, type tag and version, CBOR profile and CDDL, then fields with their own code in ascending key order; across steps the §63 order decides), with a scope paragraph for the optional steps 5, 6 and 8. - §55.1: normative trust table per section (who can write it, from which step it is bound, what it never proves); §72: security-relevant claims go in CONTROL_CBOR or under a signature, .dkk data is advisory. - §31/§54: extension arrays in strictly ascending unsigned byte order of extension_id (one rule for order and uniqueness). - Gaps a second implementation needed: §28.1 malformed age headers, §15/§19 latest unlock time and dk1_ reading rules, §22/§23/§57 length lower bounds, §63 step 8 tlock argument comparison and step 9 order, §12.1 profile validation with the drand chain-hash formula, §74 table of implementation limits. Reference alignment: .dkk errors only at step 9.a (new OpenOptions.AccessKeyFile, used by the CLI), CR/LF in dk1_ is ERR_DATEKEY_INVALID, BODY_LEN 0 is ERR_INTEGRITY, nil identities are not credentials, and AccessIdentity tries every identity on every stanza so its verdict does not depend on their order. dk1.json gains three vectors; every other testdata file is byte-identical. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks ago
seekable file, so that the `capsule_digest` of an offered `.dkk` is checked
before any release request (spec §63 step 9.a).
- `dkk`: the hex of a complete `.dkk` file (prelude and body) offered as the
access credential; absent when none is offered.
- `identities`: age X25519 identities (`AGE-SECRET-KEY-1…`) offered as access
credentials; absent when none.
- `release`: what the release source answers to every request, whatever round
is asked for. The reader must verify it (§51): a case may serve a release of
Spec v0.8.2 amendment: canonical point encoding; no library error text Amendment of the unreleased v0.8.2, recorded in §76 with its case: the second implementation's phase-2 research found that tlock-js over @noble/curves 1.9.7 accepts U re-encoded as c0 + p and a signature x + p and returns the same file key, while the reference rejects both (noble 1.9.7 differed from kilic on 5,615 of 41,686 encodings), and the spec did not say which encodings are valid. - §12.2 defines the canonical encoding of a BLS12-381 point (drand's compressed ZCash form) and requires decoders to reject every other byte string; §12.1 applies it to public_key. - §63 step 10 applies it to the release signature (ERR_RELEASE_INVALID) and step 11 defines the tlock stanza body U || V || W (96 + 16 + 16 bytes for Quicknet) with a canonical, non-infinity U (ERR_INTEGRITY). - §64 gains ten mutations, exported to mutations.json (65 cases). The signature x + p case uses published Quicknet round 1004, the first after 1000 whose x allows x + p < 2^381. The reference already gave every stated code and step. Errors no longer copy text from tlock, kyber, age, drand or kyber-bls12381. kyber's IBE error carried the candidate plaintext and r, and with one bit of W flipped the message disclosed the real tlock file key with that bit flipped. Every such place now uses a fixed reason with its normative sentinel; TestTlockFailureDiagnosticsCarryNoSecrets fails with the old wrapping. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks ago
another round, a round with the signature of another, or a signature that is
not the canonical encoding of a point (§12.2). `null` means that no release
Spec v0.8.2: corrections from the formal review A formal review of the whole v0.8.2 text found it approvable after these corrections, recorded in §76 ("Correcciones de la revisión formal"): - §27 no longer calls header_binding the authenticity of PUBLIC_HEADER: it binds the header to the opened control, never authorship or date (§55.1); the age MAC only protects against whoever lacks the file key. - §63 steps 9 and 10: a network source (relay, Release API, cache) MUST verify every response and gives ERR_RELEASE_UNAVAILABLE at step 9 when none verifies; the step-10 codes are for a directly supplied release. The reference already behaved so; TestReleaseFromANetworkSource pins both paths. - §54 and §72: registrations declare the objects and arrays where an extension may appear, and a known extension out of place counts as unknown there. The reference gains the optional extension.Placement interface, used at steps 4, 9.a and 14. - §63 step 11 fixes the GT serialization hashed by H2 (kilic/kyber order) with the frozen vector H2(e(G1, G2))[:16] = cb87319f..., shared as testdata/vectors/tlock_ibe.json; H2-H4 are cited to drand/kyber. - Step 5 makes the SEALED_CONTROL read mandatory, step 15 names ERR_HEADER_BINDING, §21 makes capsule_id 16 CSPRNG bytes a MUST, §76 is made accurate (four dk1.json vectors, the §36 time_only rule, two cases rewritten against the texts that really existed), and editorial fixes in §5, §36, §55.1, §69.1 and §77. §73 lists the three new decisions. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks ago
is available (`ERR_RELEASE_UNAVAILABLE`). The release is supplied directly,
as the caller's own, so one that breaks the rules of step 10 gets the code
of step 10; a source that fetched it over a network would have discarded
it, and the code would be `ERR_RELEASE_UNAVAILABLE` at step 9 (spec §63
steps 9 and 10).
- `now`: the reader's clock, RFC 3339. No release is requested before the round
time of the DateKey.
- `registry`: `default` pins exactly the Quicknet profile of
`profile_quicknet.json`; `empty` pins none.
- `extensions`: the extensions the application implements. Each entry is known
at `(id, version)`, and its data is valid only when it equals the bytes of
Spec v0.8.2: second-round corrections from the formal review The second round of the formal review confirmed the nine corrections of c57ed48 and asked for these, recorded in §76 as corrections 4 to 6 and an editorial note: - §72: an encoder MUST NOT write a registered extension in an object or array it is not registered for; §54: a reader MUST NOT interpret the data of a noncritical one it ignores for that reason. capsule.Encrypt and accesskey.Encode take no Registry, so the application applies the rule; their documentation and extension.Placement say so. - §17 and §51 give the step-10 codes only for a directly supplied release, as step 10 does; a network source discards a failing one at step 9. - Step 9 reports ERR_RELEASE_UNAVAILABLE and no other code, whatever the failure of the source. provider/drand.Client keeps each relay's failure as text only (errors.Join made a relay's ERR_ROUND_MISMATCH match with errors.Is), and capsule.Open keeps only the text of a source error that carries another code (a caller's source failing with ERR_RELEASE_INVALID gave that code at step 9). A context that ended stays detectable: Fetch now has a single failure path, so the canceled and deadline cases are deterministic. - TestExtensionPlacement covers the noncritical array of a .dkk: with the object-blind extension.CheckNoncritical at step 9.a it fails. - Editorial: §28.1 "analizan solo la cabecera age", one arrow at step 9, two §76 introductions; §73 lines for release sources and placement. - testdata/README.md says the corpus registers its extensions in both arrays of every object; traceability, CHANGELOG and both READMEs (integrity holds against whoever lacks the file keys, §27, §55.1) follow. Spec dated 28 September 2026; new SHA-256 in spec/README.md. No fixture or vector changes. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
`valid_data`. Each entry is registered in both arrays of every object
(§72), as a Go `extension.Registry` that does not implement
`extension.Placement`: an extension outside its registration counts as
unknown there (§54), so the cases "known critical PUBLIC_HEADER extension
with invalid data", "known critical CONTROL_CBOR extension with invalid
data" and "known critical .dkk extension with invalid data", at steps 4, 14
and 9, depend on it. Absent: the application knows no extension, the state
of the base protocol V1.
- `network`: whether the failure may come after a release request. When false,
the reader must fail without requesting any release (§27, §63): every failure
Spec v0.8.2 refinements: error precedence, trust model, strict order Approved refinements, each recorded with its reproducible case in the §76 v0.8.2 subsection: - §69.1: layered error model with normative precedence (frame, type tag and version, CBOR profile and CDDL, then fields with their own code in ascending key order; across steps the §63 order decides), with a scope paragraph for the optional steps 5, 6 and 8. - §55.1: normative trust table per section (who can write it, from which step it is bound, what it never proves); §72: security-relevant claims go in CONTROL_CBOR or under a signature, .dkk data is advisory. - §31/§54: extension arrays in strictly ascending unsigned byte order of extension_id (one rule for order and uniqueness). - Gaps a second implementation needed: §28.1 malformed age headers, §15/§19 latest unlock time and dk1_ reading rules, §22/§23/§57 length lower bounds, §63 step 8 tlock argument comparison and step 9 order, §12.1 profile validation with the drand chain-hash formula, §74 table of implementation limits. Reference alignment: .dkk errors only at step 9.a (new OpenOptions.AccessKeyFile, used by the CLI), CR/LF in dk1_ is ERR_DATEKEY_INVALID, BODY_LEN 0 is ERR_INTEGRITY, nil identities are not credentials, and AccessIdentity tries every identity on every stanza so its verdict does not depend on their order. dk1.json gains three vectors; every other testdata file is byte-identical. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks ago
of steps 1 to 8 and of step 9 before the request (9.a to 9.c).
- `frozen`: the capsule was built once with age randomness; its bytes are kept
and never regenerated. These cases have no `base`.
- `error`, `step`: the expected code and the step of §63 that fails.
Every case reproduces offline: the recorded release stands in for the network.
A reader that implements only steps 1 to 8 can replay every case whose `step` is
Implement capsule format 2 of spec v0.9 The reference moves to the DateKeys Protocol Specification v0.9, approved by its author on 29 September 2026. Encrypt writes capsule format 2 only; Open and Inspect read formats 1 and 2, and a format 1 capsule keeps the verdict v0.8.2 gave it. Format 2 (spec §22, §29.1, §31, §39): - VERSION in the PRELUDE is the capsule format, capsule.Format; any other value is ERR_UNSUPPORTED_VERSION at step 2. - CONTROL_CBOR has the schema version of its format. Version 2 adds key 6, payload_length (8 bytes, big-endian, at most L_MAX = 2^53 - 2^46), and key 7, padding (1 bloque256, 2 reforzado); it is 103 bytes without extensions, whatever L. - The payload is the content padded with zeros to P = rule(L). Step 17 checks the length and the zeros, and Open writes only the first L bytes. - INNER_ACCESS_AGE holds exactly 16 X25519 stanzas: 1 to 16 credentials, and a dummy in each slot left, in a uniformly random order. Writer rules (spec §62.1): EncryptOptions.Length is required and the source must deliver exactly that many bytes; recipients that are not canonical or of low order are rejected (agewrap.CheckX25519Recipient); self-checks of the header, the control, INNER_ACCESS_AGE and PAYLOAD_AGE. The CLI measures its input, takes -padding and reports the format. Test data: seven format 2 fixtures, padding vectors checked against math/big, format 2 CBOR vectors, and the mutation corpus in both formats with the 22 cases of the third list of spec §64, built without randomness by sealing the fixtures again with their known keys and nonces. The format 1 fixtures are kept byte for byte and never regenerated; the differential corpus keeps its 1825 cases and adds a block per format 2 fixture. The spec copy loses its "to be implemented" markers, and the READMEs, CHANGELOG, traceability and testdata/README.md follow v0.9. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
at most 8: 44 cases, 26 of them from §64. Steps 1 to 8 are summarised in "The
checks of steps 1 to 8" below.
Spec v0.8.2 refinements: error precedence, trust model, strict order Approved refinements, each recorded with its reproducible case in the §76 v0.8.2 subsection: - §69.1: layered error model with normative precedence (frame, type tag and version, CBOR profile and CDDL, then fields with their own code in ascending key order; across steps the §63 order decides), with a scope paragraph for the optional steps 5, 6 and 8. - §55.1: normative trust table per section (who can write it, from which step it is bound, what it never proves); §72: security-relevant claims go in CONTROL_CBOR or under a signature, .dkk data is advisory. - §31/§54: extension arrays in strictly ascending unsigned byte order of extension_id (one rule for order and uniqueness). - Gaps a second implementation needed: §28.1 malformed age headers, §15/§19 latest unlock time and dk1_ reading rules, §22/§23/§57 length lower bounds, §63 step 8 tlock argument comparison and step 9 order, §12.1 profile validation with the drand chain-hash formula, §74 table of implementation limits. Reference alignment: .dkk errors only at step 9.a (new OpenOptions.AccessKeyFile, used by the CLI), CR/LF in dk1_ is ERR_DATEKEY_INVALID, BODY_LEN 0 is ERR_INTEGRITY, nil identities are not credentials, and AccessIdentity tries every identity on every stanza so its verdict does not depend on their order. dk1.json gains three vectors; every other testdata file is byte-identical. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks ago
### Credentials and the release: step 9
Spec v0.8.2 refinements: error precedence, trust model, strict order Approved refinements, each recorded with its reproducible case in the §76 v0.8.2 subsection: - §69.1: layered error model with normative precedence (frame, type tag and version, CBOR profile and CDDL, then fields with their own code in ascending key order; across steps the §63 order decides), with a scope paragraph for the optional steps 5, 6 and 8. - §55.1: normative trust table per section (who can write it, from which step it is bound, what it never proves); §72: security-relevant claims go in CONTROL_CBOR or under a signature, .dkk data is advisory. - §31/§54: extension arrays in strictly ascending unsigned byte order of extension_id (one rule for order and uniqueness). - Gaps a second implementation needed: §28.1 malformed age headers, §15/§19 latest unlock time and dk1_ reading rules, §22/§23/§57 length lower bounds, §63 step 8 tlock argument comparison and step 9 order, §12.1 profile validation with the drand chain-hash formula, §74 table of implementation limits. Reference alignment: .dkk errors only at step 9.a (new OpenOptions.AccessKeyFile, used by the CLI), CR/LF in dk1_ is ERR_DATEKEY_INVALID, BODY_LEN 0 is ERR_INTEGRITY, nil identities are not credentials, and AccessIdentity tries every identity on every stanza so its verdict does not depend on their order. dk1.json gains three vectors; every other testdata file is byte-identical. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks ago
What happens between step 8 and the release request is spec §63 step 9: for
`time_and_key` only, an offered `.dkk` is checked as an object and then bound
to the capsule (9.a), at least one credential must be offered (9.b), and for
either policy a `now` before the round time of the DateKey fails without a
request (9.c); only then is the release requested. For `time_only` the
credentials play no part: the §64 case "access_policy=time_only with
time_and_key structure" offers a `.dkk` whose `capsule_digest` is that of the
unmutated capsule, and fails at step 12, not at step 9. The codes after the
request, for the release (step 10) and for the identities that open each age
file (steps 11, 13 and 17), are those of spec §63 as well. In this corpus:
Spec v0.8.2 refinements: error precedence, trust model, strict order Approved refinements, each recorded with its reproducible case in the §76 v0.8.2 subsection: - §69.1: layered error model with normative precedence (frame, type tag and version, CBOR profile and CDDL, then fields with their own code in ascending key order; across steps the §63 order decides), with a scope paragraph for the optional steps 5, 6 and 8. - §55.1: normative trust table per section (who can write it, from which step it is bound, what it never proves); §72: security-relevant claims go in CONTROL_CBOR or under a signature, .dkk data is advisory. - §31/§54: extension arrays in strictly ascending unsigned byte order of extension_id (one rule for order and uniqueness). - Gaps a second implementation needed: §28.1 malformed age headers, §15/§19 latest unlock time and dk1_ reading rules, §22/§23/§57 length lower bounds, §63 step 8 tlock argument comparison and step 9 order, §12.1 profile validation with the drand chain-hash formula, §74 table of implementation limits. Reference alignment: .dkk errors only at step 9.a (new OpenOptions.AccessKeyFile, used by the CLI), CR/LF in dk1_ is ERR_DATEKEY_INVALID, BODY_LEN 0 is ERR_INTEGRITY, nil identities are not credentials, and AccessIdentity tries every identity on every stanza so its verdict does not depend on their order. dk1.json gains three vectors; every other testdata file is byte-identical. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks ago
- identities are not examined before the release (spec §63 step 13);
- a `release` of `null` is `ERR_RELEASE_UNAVAILABLE` at step 9;
Spec v0.8.2: corrections from the formal review A formal review of the whole v0.8.2 text found it approvable after these corrections, recorded in §76 ("Correcciones de la revisión formal"): - §27 no longer calls header_binding the authenticity of PUBLIC_HEADER: it binds the header to the opened control, never authorship or date (§55.1); the age MAC only protects against whoever lacks the file key. - §63 steps 9 and 10: a network source (relay, Release API, cache) MUST verify every response and gives ERR_RELEASE_UNAVAILABLE at step 9 when none verifies; the step-10 codes are for a directly supplied release. The reference already behaved so; TestReleaseFromANetworkSource pins both paths. - §54 and §72: registrations declare the objects and arrays where an extension may appear, and a known extension out of place counts as unknown there. The reference gains the optional extension.Placement interface, used at steps 4, 9.a and 14. - §63 step 11 fixes the GT serialization hashed by H2 (kilic/kyber order) with the frozen vector H2(e(G1, G2))[:16] = cb87319f..., shared as testdata/vectors/tlock_ibe.json; H2-H4 are cited to drand/kyber. - Step 5 makes the SEALED_CONTROL read mandatory, step 15 names ERR_HEADER_BINDING, §21 makes capsule_id 16 CSPRNG bytes a MUST, §76 is made accurate (four dk1.json vectors, the §36 time_only rule, two cases rewritten against the texts that really existed), and editorial fixes in §5, §36, §55.1, §69.1 and §77. §73 lists the three new decisions. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks ago
- every `release` is supplied directly, so an invalid one fails at step 10:
the source is not a network source, which would discard it at step 9;
Spec v0.8.2 refinements: error precedence, trust model, strict order Approved refinements, each recorded with its reproducible case in the §76 v0.8.2 subsection: - §69.1: layered error model with normative precedence (frame, type tag and version, CBOR profile and CDDL, then fields with their own code in ascending key order; across steps the §63 order decides), with a scope paragraph for the optional steps 5, 6 and 8. - §55.1: normative trust table per section (who can write it, from which step it is bound, what it never proves); §72: security-relevant claims go in CONTROL_CBOR or under a signature, .dkk data is advisory. - §31/§54: extension arrays in strictly ascending unsigned byte order of extension_id (one rule for order and uniqueness). - Gaps a second implementation needed: §28.1 malformed age headers, §15/§19 latest unlock time and dk1_ reading rules, §22/§23/§57 length lower bounds, §63 step 8 tlock argument comparison and step 9 order, §12.1 profile validation with the drand chain-hash formula, §74 table of implementation limits. Reference alignment: .dkk errors only at step 9.a (new OpenOptions.AccessKeyFile, used by the CLI), CR/LF in dk1_ is ERR_DATEKEY_INVALID, BODY_LEN 0 is ERR_INTEGRITY, nil identities are not credentials, and AccessIdentity tries every identity on every stanza so its verdict does not depend on their order. dk1.json gains three vectors; every other testdata file is byte-identical. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks ago
- every `.dkk` offered decodes: the corpus checks step 9.a, not the decoding
of a `.dkk`, whose errors spec §63 also places at step 9.a.
Spec v0.8.2 amendment: canonical point encoding; no library error text Amendment of the unreleased v0.8.2, recorded in §76 with its case: the second implementation's phase-2 research found that tlock-js over @noble/curves 1.9.7 accepts U re-encoded as c0 + p and a signature x + p and returns the same file key, while the reference rejects both (noble 1.9.7 differed from kilic on 5,615 of 41,686 encodings), and the spec did not say which encodings are valid. - §12.2 defines the canonical encoding of a BLS12-381 point (drand's compressed ZCash form) and requires decoders to reject every other byte string; §12.1 applies it to public_key. - §63 step 10 applies it to the release signature (ERR_RELEASE_INVALID) and step 11 defines the tlock stanza body U || V || W (96 + 16 + 16 bytes for Quicknet) with a canonical, non-infinity U (ERR_INTEGRITY). - §64 gains ten mutations, exported to mutations.json (65 cases). The signature x + p case uses published Quicknet round 1004, the first after 1000 whose x allows x + p < 2^381. The reference already gave every stated code and step. Errors no longer copy text from tlock, kyber, age, drand or kyber-bls12381. kyber's IBE error carried the candidate plaintext and r, and with one bit of W flipped the message disclosed the real tlock file key with that bit flipped. Every such place now uses a fixed reason with its normative sentinel; TestTlockFailureDiagnosticsCarryNoSecrets fails with the old wrapping. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks ago
### Point encodings: steps 10 and 11
Implement capsule format 2 of spec v0.9 The reference moves to the DateKeys Protocol Specification v0.9, approved by its author on 29 September 2026. Encrypt writes capsule format 2 only; Open and Inspect read formats 1 and 2, and a format 1 capsule keeps the verdict v0.8.2 gave it. Format 2 (spec §22, §29.1, §31, §39): - VERSION in the PRELUDE is the capsule format, capsule.Format; any other value is ERR_UNSUPPORTED_VERSION at step 2. - CONTROL_CBOR has the schema version of its format. Version 2 adds key 6, payload_length (8 bytes, big-endian, at most L_MAX = 2^53 - 2^46), and key 7, padding (1 bloque256, 2 reforzado); it is 103 bytes without extensions, whatever L. - The payload is the content padded with zeros to P = rule(L). Step 17 checks the length and the zeros, and Open writes only the first L bytes. - INNER_ACCESS_AGE holds exactly 16 X25519 stanzas: 1 to 16 credentials, and a dummy in each slot left, in a uniformly random order. Writer rules (spec §62.1): EncryptOptions.Length is required and the source must deliver exactly that many bytes; recipients that are not canonical or of low order are rejected (agewrap.CheckX25519Recipient); self-checks of the header, the control, INNER_ACCESS_AGE and PAYLOAD_AGE. The CLI measures its input, takes -padding and reports the format. Test data: seven format 2 fixtures, padding vectors checked against math/big, format 2 CBOR vectors, and the mutation corpus in both formats with the 22 cases of the third list of spec §64, built without randomness by sealing the fixtures again with their known keys and nonces. The format 1 fixtures are kept byte for byte and never regenerated; the differential corpus keeps its 1825 cases and adds a block per format 2 fixture. The spec copy loses its "to be implemented" markers, and the READMEs, CHANGELOG, traceability and testdata/README.md follow v0.9. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
Ten cases of §64, in each format, test the canonical point encoding of spec
§12.2 and the tlock stanza body `U || V || W` of spec §63 step 11:
Spec v0.8.2 amendment: canonical point encoding; no library error text Amendment of the unreleased v0.8.2, recorded in §76 with its case: the second implementation's phase-2 research found that tlock-js over @noble/curves 1.9.7 accepts U re-encoded as c0 + p and a signature x + p and returns the same file key, while the reference rejects both (noble 1.9.7 differed from kilic on 5,615 of 41,686 encodings), and the spec did not say which encodings are valid. - §12.2 defines the canonical encoding of a BLS12-381 point (drand's compressed ZCash form) and requires decoders to reject every other byte string; §12.1 applies it to public_key. - §63 step 10 applies it to the release signature (ERR_RELEASE_INVALID) and step 11 defines the tlock stanza body U || V || W (96 + 16 + 16 bytes for Quicknet) with a canonical, non-infinity U (ERR_INTEGRITY). - §64 gains ten mutations, exported to mutations.json (65 cases). The signature x + p case uses published Quicknet round 1004, the first after 1000 whose x allows x + p < 2^381. The reference already gave every stated code and step. Errors no longer copy text from tlock, kyber, age, drand or kyber-bls12381. kyber's IBE error carried the candidate plaintext and r, and with one bit of W flipped the message disclosed the real tlock file key with that bit flipped. Every such place now uses a fixed reason with its normative sentinel; TestTlockFailureDiagnosticsCarryNoSecrets fails with the old wrapping. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks ago
- five edit the tlock stanza body of `time_only.dkc`: U with c0 + p, U the
point at infinity (the byte 0xc0, then zeros), U with the infinity flag
over its own coordinate bits, and bodies of 127 and 129 bytes. Each recomputes the header
MAC of OUTER_TIME_AGE with FK_TIME, so the age header stays authentic and
only the rules of step 11 reject the capsule (`ERR_INTEGRITY`); a reader
whose decoder reduces coordinates modulo p opens the first one;
- three serve the release of `time_only.dkc` with its signature edited: the
point at infinity, the infinity flag over its own coordinate bits, and the
negated signature (the sort flag flipped: a canonical point that does not
verify); a fourth serves the published signature of round 1004 re-encoded
as x + p, over a frozen capsule for round 1004, because no fixture round
has a signature whose x + p fits in 381 bits. All four fail at step 10
(`ERR_RELEASE_INVALID`);
- the last one serves the negated signature with U with c0 + p: step 10
comes first.
Implement capsule format 2 of spec v0.9 The reference moves to the DateKeys Protocol Specification v0.9, approved by its author on 29 September 2026. Encrypt writes capsule format 2 only; Open and Inspect read formats 1 and 2, and a format 1 capsule keeps the verdict v0.8.2 gave it. Format 2 (spec §22, §29.1, §31, §39): - VERSION in the PRELUDE is the capsule format, capsule.Format; any other value is ERR_UNSUPPORTED_VERSION at step 2. - CONTROL_CBOR has the schema version of its format. Version 2 adds key 6, payload_length (8 bytes, big-endian, at most L_MAX = 2^53 - 2^46), and key 7, padding (1 bloque256, 2 reforzado); it is 103 bytes without extensions, whatever L. - The payload is the content padded with zeros to P = rule(L). Step 17 checks the length and the zeros, and Open writes only the first L bytes. - INNER_ACCESS_AGE holds exactly 16 X25519 stanzas: 1 to 16 credentials, and a dummy in each slot left, in a uniformly random order. Writer rules (spec §62.1): EncryptOptions.Length is required and the source must deliver exactly that many bytes; recipients that are not canonical or of low order are rejected (agewrap.CheckX25519Recipient); self-checks of the header, the control, INNER_ACCESS_AGE and PAYLOAD_AGE. The CLI measures its input, takes -padding and reports the format. Test data: seven format 2 fixtures, padding vectors checked against math/big, format 2 CBOR vectors, and the mutation corpus in both formats with the 22 cases of the third list of spec §64, built without randomness by sealing the fixtures again with their known keys and nonces. The format 1 fixtures are kept byte for byte and never regenerated; the differential corpus keeps its 1825 cases and adds a block per format 2 fixture. The spec copy loses its "to be implemented" markers, and the READMEs, CHANGELOG, traceability and testdata/README.md follow v0.9. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
In format 2 the same ten edit `format2_time_only.dkc` and its release, and the
x + p case has a frozen format 2 capsule for round 1004.
### Format 2: the third list of §64
Format 3, step 3: the reader Open reads format 3 (spec 29.2 to 29.7, 63 steps 17 and 18): the PRELUDE accepts VERSION 3, and the files go to a Sink. - Sink: Begin with the validated head, Create for each file in the order of the head, and Commit only after every check of step 17; after any failure that follows a successful Begin, Abort, once. A format 3 capsule without a Sink fails right after step 2 with ErrSinkRequired, a caller error with no code, no failed step and no request; a capsule of format 1 or 2 without dst fails there too. - Step 17 in its substeps: the frame and the area, security and its verdicts, which never fail, the head, the files filling CONTENT, the SHA-256 of each file and the padding. A failure of age or a plaintext whose length is not P prevails; otherwise the first substep that fails decides, and a code other than ERR_INTEGRITY is reported only after reading PAYLOAD_AGE to its end. - The reads of BODY grow with the bytes received, never with AREA_LEN, HEAD_LEN or a declared size (spec 57); a test measures it. - A failure of the Sink is the caller's own error with ERR_INTEGRITY, as one of dst is in formats 1 and 2. - Opened gains Head, Verdicts, AreaLen and UnusableHeadExtensions. - Test data: "version changed" sets VERSION 4, and the format 2 list gains "format 2 time_only relabeled format 3", which fails at step 14, as section 64 of spec v0.10 lists: 126 cases, 89 of the spec. The randomly built capsules keep their recorded bytes. - testkit: Build writes format 3 and can edit the padded plaintext; Head3, Body3, DiscardSink and MemorySink build and open BODY. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
The 23 cases of the third list of §64 test what format 2 adds:
Implement capsule format 2 of spec v0.9 The reference moves to the DateKeys Protocol Specification v0.9, approved by its author on 29 September 2026. Encrypt writes capsule format 2 only; Open and Inspect read formats 1 and 2, and a format 1 capsule keeps the verdict v0.8.2 gave it. Format 2 (spec §22, §29.1, §31, §39): - VERSION in the PRELUDE is the capsule format, capsule.Format; any other value is ERR_UNSUPPORTED_VERSION at step 2. - CONTROL_CBOR has the schema version of its format. Version 2 adds key 6, payload_length (8 bytes, big-endian, at most L_MAX = 2^53 - 2^46), and key 7, padding (1 bloque256, 2 reforzado); it is 103 bytes without extensions, whatever L. - The payload is the content padded with zeros to P = rule(L). Step 17 checks the length and the zeros, and Open writes only the first L bytes. - INNER_ACCESS_AGE holds exactly 16 X25519 stanzas: 1 to 16 credentials, and a dummy in each slot left, in a uniformly random order. Writer rules (spec §62.1): EncryptOptions.Length is required and the source must deliver exactly that many bytes; recipients that are not canonical or of low order are rejected (agewrap.CheckX25519Recipient); self-checks of the header, the control, INNER_ACCESS_AGE and PAYLOAD_AGE. The CLI measures its input, takes -padding and reports the format. Test data: seven format 2 fixtures, padding vectors checked against math/big, format 2 CBOR vectors, and the mutation corpus in both formats with the 22 cases of the third list of spec §64, built without randomness by sealing the fixtures again with their known keys and nonces. The format 1 fixtures are kept byte for byte and never regenerated; the differential corpus keeps its 1825 cases and adds a block per format 2 fixture. The spec copy loses its "to be implemented" markers, and the READMEs, CHANGELOG, traceability and testdata/README.md follow v0.9. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
Format 3, step 3: the reader Open reads format 3 (spec 29.2 to 29.7, 63 steps 17 and 18): the PRELUDE accepts VERSION 3, and the files go to a Sink. - Sink: Begin with the validated head, Create for each file in the order of the head, and Commit only after every check of step 17; after any failure that follows a successful Begin, Abort, once. A format 3 capsule without a Sink fails right after step 2 with ErrSinkRequired, a caller error with no code, no failed step and no request; a capsule of format 1 or 2 without dst fails there too. - Step 17 in its substeps: the frame and the area, security and its verdicts, which never fail, the head, the files filling CONTENT, the SHA-256 of each file and the padding. A failure of age or a plaintext whose length is not P prevails; otherwise the first substep that fails decides, and a code other than ERR_INTEGRITY is reported only after reading PAYLOAD_AGE to its end. - The reads of BODY grow with the bytes received, never with AREA_LEN, HEAD_LEN or a declared size (spec 57); a test measures it. - A failure of the Sink is the caller's own error with ERR_INTEGRITY, as one of dst is in formats 1 and 2. - Opened gains Head, Verdicts, AreaLen and UnusableHeadExtensions. - Test data: "version changed" sets VERSION 4, and the format 2 list gains "format 2 time_only relabeled format 3", which fails at step 14, as section 64 of spec v0.10 lists: 126 cases, 89 of the spec. The randomly built capsules keep their recorded bytes. - testkit: Build writes format 3 and can edit the padded plaintext; Head3, Body3, DiscardSink and MemorySink build and open BODY. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
- the format against the rest of the capsule, with another `VERSION`, a
Implement capsule format 2 of spec v0.9 The reference moves to the DateKeys Protocol Specification v0.9, approved by its author on 29 September 2026. Encrypt writes capsule format 2 only; Open and Inspect read formats 1 and 2, and a format 1 capsule keeps the verdict v0.8.2 gave it. Format 2 (spec §22, §29.1, §31, §39): - VERSION in the PRELUDE is the capsule format, capsule.Format; any other value is ERR_UNSUPPORTED_VERSION at step 2. - CONTROL_CBOR has the schema version of its format. Version 2 adds key 6, payload_length (8 bytes, big-endian, at most L_MAX = 2^53 - 2^46), and key 7, padding (1 bloque256, 2 reforzado); it is 103 bytes without extensions, whatever L. - The payload is the content padded with zeros to P = rule(L). Step 17 checks the length and the zeros, and Open writes only the first L bytes. - INNER_ACCESS_AGE holds exactly 16 X25519 stanzas: 1 to 16 credentials, and a dummy in each slot left, in a uniformly random order. Writer rules (spec §62.1): EncryptOptions.Length is required and the source must deliver exactly that many bytes; recipients that are not canonical or of low order are rejected (agewrap.CheckX25519Recipient); self-checks of the header, the control, INNER_ACCESS_AGE and PAYLOAD_AGE. The CLI measures its input, takes -padding and reports the format. Test data: seven format 2 fixtures, padding vectors checked against math/big, format 2 CBOR vectors, and the mutation corpus in both formats with the 22 cases of the third list of spec §64, built without randomness by sealing the fixtures again with their known keys and nonces. The format 1 fixtures are kept byte for byte and never regenerated; the differential corpus keeps its 1825 cases and adds a block per format 2 fixture. The spec copy loses its "to be implemented" markers, and the READMEs, CHANGELOG, traceability and testdata/README.md follow v0.9. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
public byte (§22, §76): a format 1 `time_and_key` capsule of one stanza
Format 3, step 3: the reader Open reads format 3 (spec 29.2 to 29.7, 63 steps 17 and 18): the PRELUDE accepts VERSION 3, and the files go to a Sink. - Sink: Begin with the validated head, Create for each file in the order of the head, and Commit only after every check of step 17; after any failure that follows a successful Begin, Abort, once. A format 3 capsule without a Sink fails right after step 2 with ErrSinkRequired, a caller error with no code, no failed step and no request; a capsule of format 1 or 2 without dst fails there too. - Step 17 in its substeps: the frame and the area, security and its verdicts, which never fail, the head, the files filling CONTENT, the SHA-256 of each file and the padding. A failure of age or a plaintext whose length is not P prevails; otherwise the first substep that fails decides, and a code other than ERR_INTEGRITY is reported only after reading PAYLOAD_AGE to its end. - The reads of BODY grow with the bytes received, never with AREA_LEN, HEAD_LEN or a declared size (spec 57); a test measures it. - A failure of the Sink is the caller's own error with ERR_INTEGRITY, as one of dst is in formats 1 and 2. - Opened gains Head, Verdicts, AreaLen and UnusableHeadExtensions. - Test data: "version changed" sets VERSION 4, and the format 2 list gains "format 2 time_only relabeled format 3", which fails at step 14, as section 64 of spec v0.10 lists: 126 cases, 89 of the spec. The randomly built capsules keep their recorded bytes. - testkit: Build writes format 3 and can edit the padded plaintext; Head3, Body3, DiscardSink and MemorySink build and open BODY. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
relabeled format 2 fails at step 12, and the other relabelings, format 2 as
1 or 3 and format 1 as 2, at step 14, where the schema version of the
control is not the format;
Implement capsule format 2 of spec v0.9 The reference moves to the DateKeys Protocol Specification v0.9, approved by its author on 29 September 2026. Encrypt writes capsule format 2 only; Open and Inspect read formats 1 and 2, and a format 1 capsule keeps the verdict v0.8.2 gave it. Format 2 (spec §22, §29.1, §31, §39): - VERSION in the PRELUDE is the capsule format, capsule.Format; any other value is ERR_UNSUPPORTED_VERSION at step 2. - CONTROL_CBOR has the schema version of its format. Version 2 adds key 6, payload_length (8 bytes, big-endian, at most L_MAX = 2^53 - 2^46), and key 7, padding (1 bloque256, 2 reforzado); it is 103 bytes without extensions, whatever L. - The payload is the content padded with zeros to P = rule(L). Step 17 checks the length and the zeros, and Open writes only the first L bytes. - INNER_ACCESS_AGE holds exactly 16 X25519 stanzas: 1 to 16 credentials, and a dummy in each slot left, in a uniformly random order. Writer rules (spec §62.1): EncryptOptions.Length is required and the source must deliver exactly that many bytes; recipients that are not canonical or of low order are rejected (agewrap.CheckX25519Recipient); self-checks of the header, the control, INNER_ACCESS_AGE and PAYLOAD_AGE. The CLI measures its input, takes -padding and reports the format. Test data: seven format 2 fixtures, padding vectors checked against math/big, format 2 CBOR vectors, and the mutation corpus in both formats with the 22 cases of the third list of spec §64, built without randomness by sealing the fixtures again with their known keys and nonces. The format 1 fixtures are kept byte for byte and never regenerated; the differential corpus keeps its 1825 cases and adds a block per format 2 fixture. The spec copy loses its "to be implemented" markers, and the READMEs, CHANGELOG, traceability and testdata/README.md follow v0.9. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
- the 16 stanzas of INNER_ACCESS_AGE: 15, 17, two for one recipient, and an
identity that opens none of them, at steps 12 and 13;
- keys 6 and 7 of CONTROL_CBOR, at step 14;
- the plaintext of PAYLOAD_AGE, at step 17: a padding byte that is not zero,
a plaintext of P − 1 or P + 256 bytes or without its padding, and a control
that declares another code or another L.
Each derives from a fixture without randomness. What it changes is sealed again
with the file keys that the release and the credentials of the fixture open,
FK_TIME and FK_ACCESS, or that its I_PAYLOAD unwraps, FK_PAYLOAD, and with the
nonces of the fixture; a stanza it adds takes its ephemeral scalar from a fixed
seed. So each case is a `base` with edits, and regenerates byte for byte. A
case that changes a `time_and_key` capsule offers the identity of its `.dkk`,
because the `.dkk` itself would fail at step 9.a: its `capsule_digest` no
longer matches. The last five cases of the corpus offer the `.dkk` and show
that.
## The checks of steps 1 to 8
Spec v0.8.2 refinements: error precedence, trust model, strict order Approved refinements, each recorded with its reproducible case in the §76 v0.8.2 subsection: - §69.1: layered error model with normative precedence (frame, type tag and version, CBOR profile and CDDL, then fields with their own code in ascending key order; across steps the §63 order decides), with a scope paragraph for the optional steps 5, 6 and 8. - §55.1: normative trust table per section (who can write it, from which step it is bound, what it never proves); §72: security-relevant claims go in CONTROL_CBOR or under a signature, .dkk data is advisory. - §31/§54: extension arrays in strictly ascending unsigned byte order of extension_id (one rule for order and uniqueness). - Gaps a second implementation needed: §28.1 malformed age headers, §15/§19 latest unlock time and dk1_ reading rules, §22/§23/§57 length lower bounds, §63 step 8 tlock argument comparison and step 9 order, §12.1 profile validation with the drand chain-hash formula, §74 table of implementation limits. Reference alignment: .dkk errors only at step 9.a (new OpenOptions.AccessKeyFile, used by the CLI), CR/LF in dk1_ is ERR_DATEKEY_INVALID, BODY_LEN 0 is ERR_INTEGRITY, nil identities are not credentials, and AccessIdentity tries every identity on every stanza so its verdict does not depend on their order. dk1.json gains three vectors; every other testdata file is byte-identical. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks ago
`mutations.json` and `inspect_differential.json` follow the rules of the
specification for steps 1 to 8, with the `default` registry (Quicknet pinned),
no extension known unless a case lists some, and no secret. The first failure
ends the flow, and within one object the first failing layer decides the code
(spec §69.1):
Spec v0.8.2 refinements: error precedence, trust model, strict order Approved refinements, each recorded with its reproducible case in the §76 v0.8.2 subsection: - §69.1: layered error model with normative precedence (frame, type tag and version, CBOR profile and CDDL, then fields with their own code in ascending key order; across steps the §63 order decides), with a scope paragraph for the optional steps 5, 6 and 8. - §55.1: normative trust table per section (who can write it, from which step it is bound, what it never proves); §72: security-relevant claims go in CONTROL_CBOR or under a signature, .dkk data is advisory. - §31/§54: extension arrays in strictly ascending unsigned byte order of extension_id (one rule for order and uniqueness). - Gaps a second implementation needed: §28.1 malformed age headers, §15/§19 latest unlock time and dk1_ reading rules, §22/§23/§57 length lower bounds, §63 step 8 tlock argument comparison and step 9 order, §12.1 profile validation with the drand chain-hash formula, §74 table of implementation limits. Reference alignment: .dkk errors only at step 9.a (new OpenOptions.AccessKeyFile, used by the CLI), CR/LF in dk1_ is ERR_DATEKEY_INVALID, BODY_LEN 0 is ERR_INTEGRITY, nil identities are not credentials, and AccessIdentity tries every identity on every stanza so its verdict does not depend on their order. dk1.json gains three vectors; every other testdata file is byte-identical. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks ago
| Step | Rules | Spec |
|---|---|---|
Format 3, step 3: the reader Open reads format 3 (spec 29.2 to 29.7, 63 steps 17 and 18): the PRELUDE accepts VERSION 3, and the files go to a Sink. - Sink: Begin with the validated head, Create for each file in the order of the head, and Commit only after every check of step 17; after any failure that follows a successful Begin, Abort, once. A format 3 capsule without a Sink fails right after step 2 with ErrSinkRequired, a caller error with no code, no failed step and no request; a capsule of format 1 or 2 without dst fails there too. - Step 17 in its substeps: the frame and the area, security and its verdicts, which never fail, the head, the files filling CONTENT, the SHA-256 of each file and the padding. A failure of age or a plaintext whose length is not P prevails; otherwise the first substep that fails decides, and a code other than ERR_INTEGRITY is reported only after reading PAYLOAD_AGE to its end. - The reads of BODY grow with the bytes received, never with AREA_LEN, HEAD_LEN or a declared size (spec 57); a test measures it. - A failure of the Sink is the caller's own error with ERR_INTEGRITY, as one of dst is in formats 1 and 2. - Opened gains Head, Verdicts, AreaLen and UnusableHeadExtensions. - Test data: "version changed" sets VERSION 4, and the format 2 list gains "format 2 time_only relabeled format 3", which fails at step 14, as section 64 of spec v0.10 lists: 126 cases, 89 of the spec. The randomly built capsules keep their recorded bytes. - testkit: Build writes format 3 and can edit the padded plaintext; Head3, Body3, DiscardSink and MemorySink build and open BODY. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
| 1, 2 | magic, truncated prelude, `VERSION` 1, 2 or 3 (the format), FLAGS and RESERVED, `PUBLIC_HEADER_LEN` in 1 to 1048576 and `SEALED_CONTROL_LEN` in 1 to 67108864 | §22, §23 |
Spec v0.8.2 refinements: error precedence, trust model, strict order Approved refinements, each recorded with its reproducible case in the §76 v0.8.2 subsection: - §69.1: layered error model with normative precedence (frame, type tag and version, CBOR profile and CDDL, then fields with their own code in ascending key order; across steps the §63 order decides), with a scope paragraph for the optional steps 5, 6 and 8. - §55.1: normative trust table per section (who can write it, from which step it is bound, what it never proves); §72: security-relevant claims go in CONTROL_CBOR or under a signature, .dkk data is advisory. - §31/§54: extension arrays in strictly ascending unsigned byte order of extension_id (one rule for order and uniqueness). - Gaps a second implementation needed: §28.1 malformed age headers, §15/§19 latest unlock time and dk1_ reading rules, §22/§23/§57 length lower bounds, §63 step 8 tlock argument comparison and step 9 order, §12.1 profile validation with the drand chain-hash formula, §74 table of implementation limits. Reference alignment: .dkk errors only at step 9.a (new OpenOptions.AccessKeyFile, used by the CLI), CR/LF in dk1_ is ERR_DATEKEY_INVALID, BODY_LEN 0 is ERR_INTEGRITY, nil identities are not credentials, and AccessIdentity tries every identity on every stanza so its verdict does not depend on their order. dk1.json gains three vectors; every other testdata file is byte-identical. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks ago
| 3 | the PUBLIC_HEADER bytes are present | §23 |
| 4 | PUBLIC_HEADER: layers 2 to 4, the `public_header` decoder of the schema vectors, then the pinned profile of the DateKey and the critical extensions; with no extension known, every `critical_extensions` array fails | §63, §69.1 |
| 5 | the SEALED_CONTROL bytes are present, then its age header, parsed within `SEALED_CONTROL_LEN` bytes: malformed is `ERR_INTEGRITY`, one tlock stanza is required | §28.1 |
| 6 | the age header of PAYLOAD_AGE, from its offset to the end of the file: malformed is `ERR_INTEGRITY`, one X25519 stanza is required | §22, §28.1 |
| 7 | the round time of the DateKey is at most 9999-12-31T23:59:59Z (Quicknet: round 83903165811 at most); a `dk1_` round above it passes step 4 and fails here | §15 |
| 8 | the tlock stanza has exactly two arguments, the canonical decimal round and the lowercase hex chain hash, compared as strings | §63 step 8 |
The reference parses age headers with `filippo.io/age` v1.3.2, whose parser
limits (1024 stanzas, 128 arguments after the type, 2 MiB) spec §74 lists as
implementation limits. No case of these corpora depends on them.
## `vectors/inspect_differential.json`
Implement capsule format 2 of spec v0.9 The reference moves to the DateKeys Protocol Specification v0.9, approved by its author on 29 September 2026. Encrypt writes capsule format 2 only; Open and Inspect read formats 1 and 2, and a format 1 capsule keeps the verdict v0.8.2 gave it. Format 2 (spec §22, §29.1, §31, §39): - VERSION in the PRELUDE is the capsule format, capsule.Format; any other value is ERR_UNSUPPORTED_VERSION at step 2. - CONTROL_CBOR has the schema version of its format. Version 2 adds key 6, payload_length (8 bytes, big-endian, at most L_MAX = 2^53 - 2^46), and key 7, padding (1 bloque256, 2 reforzado); it is 103 bytes without extensions, whatever L. - The payload is the content padded with zeros to P = rule(L). Step 17 checks the length and the zeros, and Open writes only the first L bytes. - INNER_ACCESS_AGE holds exactly 16 X25519 stanzas: 1 to 16 credentials, and a dummy in each slot left, in a uniformly random order. Writer rules (spec §62.1): EncryptOptions.Length is required and the source must deliver exactly that many bytes; recipients that are not canonical or of low order are rejected (agewrap.CheckX25519Recipient); self-checks of the header, the control, INNER_ACCESS_AGE and PAYLOAD_AGE. The CLI measures its input, takes -padding and reports the format. Test data: seven format 2 fixtures, padding vectors checked against math/big, format 2 CBOR vectors, and the mutation corpus in both formats with the 22 cases of the third list of spec §64, built without randomness by sealing the fixtures again with their known keys and nonces. The format 1 fixtures are kept byte for byte and never regenerated; the differential corpus keeps its 1825 cases and adds a block per format 2 fixture. The spec copy loses its "to be implemented" markers, and the READMEs, CHANGELOG, traceability and testdata/README.md follow v0.9. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
A differential corpus of the pre-unlock checks: 4380 deterministic mutations of
the twelve official `.dkc` fixtures, 365 of each, with the verdict of steps 1
to 8 of §63 as the reference computes it (`capsule.Inspect` with the `default`
registry, no extension known, no network, no secret), by the rules that "The
checks of steps 1 to 8" above points to. The first 1825, those of the five
format 1 fixtures, are the corpus of v0.8.2 unchanged; the seven format 2
fixtures follow. The `format` field of the file describes the layout below in
words; it is not a capsule format.
```json
{
"seed": 20260925,
"bases": [ { "file": "time_only.dkc", "sha256": "99e9…" } ],
"mutations": [
{"base":0,"kind":"flip","edits":[[121,1,"b3"]],"result":"ERR_NON_CANONICAL_CBOR","step":4},
{"base":0,"kind":"flip","edits":[[725,1,"4d"]],"result":"ok"}
]
}
```
- `bases`: the fixtures, with the SHA-256 of their exact bytes. `base` in a
mutation is an index into this list.
- `edits`: see "Edited files".
- `result`: `ok` when steps 1 to 8 pass, or the error code; `step` is the step
that failed, absent when `ok`.
- `kind` names the generator and is informative: `flip` (one bit), `byte` (one
byte replaced), `truncate`, `insert` and `delete` (one to four bytes),
`length` (PUBLIC_HEADER_LEN and SEALED_CONTROL_LEN), `header` (PUBLIC_HEADER
re-encoded with one CBOR-aware change: a key removed, added or retyped, the
type tag, version, `capsule_id`, DateKey or `access_policy` changed, an
extension array added, a head not in its shortest form, keys out of order or
repeated, the whole item tagged, wrapped, made indefinite, truncated or
followed by bytes), `datekey` (the DateKey string alone), `age` (the age
header of SEALED_CONTROL or PAYLOAD_AGE edited: intro line, stanza type,
arguments, stanzas added or removed, body lines, MAC line, line endings).
Most `header`, `datekey` and SEALED_CONTROL `age` mutations also rewrite the
prelude lengths to match; some keep the old ones on purpose.
- `seed` seeds the generator of the reference and is informative too: every
mutation is stored explicitly.
## `fixtures/<name>.inspect.json`
For each official `.dkc`, the exact bytes that `datekeys inspect -json -in
<name>.dkc` prints when run in `testdata/fixtures`: JSON indented with two
spaces, fields in this order, and a final newline. The pre-unlock checks use
the `default` registry.
| Field | Content |
|---|---|
| `file` | the `-in` argument, `<name>.dkc` |
Implement capsule format 2 of spec v0.9 The reference moves to the DateKeys Protocol Specification v0.9, approved by its author on 29 September 2026. Encrypt writes capsule format 2 only; Open and Inspect read formats 1 and 2, and a format 1 capsule keeps the verdict v0.8.2 gave it. Format 2 (spec §22, §29.1, §31, §39): - VERSION in the PRELUDE is the capsule format, capsule.Format; any other value is ERR_UNSUPPORTED_VERSION at step 2. - CONTROL_CBOR has the schema version of its format. Version 2 adds key 6, payload_length (8 bytes, big-endian, at most L_MAX = 2^53 - 2^46), and key 7, padding (1 bloque256, 2 reforzado); it is 103 bytes without extensions, whatever L. - The payload is the content padded with zeros to P = rule(L). Step 17 checks the length and the zeros, and Open writes only the first L bytes. - INNER_ACCESS_AGE holds exactly 16 X25519 stanzas: 1 to 16 credentials, and a dummy in each slot left, in a uniformly random order. Writer rules (spec §62.1): EncryptOptions.Length is required and the source must deliver exactly that many bytes; recipients that are not canonical or of low order are rejected (agewrap.CheckX25519Recipient); self-checks of the header, the control, INNER_ACCESS_AGE and PAYLOAD_AGE. The CLI measures its input, takes -padding and reports the format. Test data: seven format 2 fixtures, padding vectors checked against math/big, format 2 CBOR vectors, and the mutation corpus in both formats with the 22 cases of the third list of spec §64, built without randomness by sealing the fixtures again with their known keys and nonces. The format 1 fixtures are kept byte for byte and never regenerated; the differential corpus keeps its 1825 cases and adds a block per format 2 fixture. The spec copy loses its "to be implemented" markers, and the READMEs, CHANGELOG, traceability and testdata/README.md follow v0.9. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
| `format` | the capsule format, once step 2 has passed |
| `capsule_id` | hex, once step 4 has decoded the header |
| `datekey`, `profile`, `round` | the canonical `dk1_` string, its profile and round |
| `unlock_at` | the round time of the DateKey, RFC 3339 in UTC, once step 7 passes |
| `access_policy` | `time_only` or `time_and_key` |
| `valid` | true when steps 1 to 8 pass |
| `error` | the code of the failure, absent when valid |
| `checks` | one entry per step run: `step`, `name`, `ok`, `detail` (free text) and, for a failed step, `error` |
`detail` is informative text of the reference; a second implementation compares
at least `step`, `name`, `ok` and `error`, and every other field.
## Existing vectors and fixtures
- `vectors/profile_quicknet.json`: the Quicknet profile fields,
`canonical_cbor` (hex) and `profile_hash`.
- `vectors/quicknet_rounds.json`: `vectors` of `requested` instants (RFC 3339
with nanoseconds) and the resolved `round` and `effective` time, or `error`.
- `vectors/dk1.json`: valid `vectors` with `network`, `round`,
`canonical_json`, `base64url` and `dk1`; invalid ones with `input` and the
`error` code (`accepted` would mean the input decodes).
Implement capsule format 2 of spec v0.9 The reference moves to the DateKeys Protocol Specification v0.9, approved by its author on 29 September 2026. Encrypt writes capsule format 2 only; Open and Inspect read formats 1 and 2, and a format 1 capsule keeps the verdict v0.8.2 gave it. Format 2 (spec §22, §29.1, §31, §39): - VERSION in the PRELUDE is the capsule format, capsule.Format; any other value is ERR_UNSUPPORTED_VERSION at step 2. - CONTROL_CBOR has the schema version of its format. Version 2 adds key 6, payload_length (8 bytes, big-endian, at most L_MAX = 2^53 - 2^46), and key 7, padding (1 bloque256, 2 reforzado); it is 103 bytes without extensions, whatever L. - The payload is the content padded with zeros to P = rule(L). Step 17 checks the length and the zeros, and Open writes only the first L bytes. - INNER_ACCESS_AGE holds exactly 16 X25519 stanzas: 1 to 16 credentials, and a dummy in each slot left, in a uniformly random order. Writer rules (spec §62.1): EncryptOptions.Length is required and the source must deliver exactly that many bytes; recipients that are not canonical or of low order are rejected (agewrap.CheckX25519Recipient); self-checks of the header, the control, INNER_ACCESS_AGE and PAYLOAD_AGE. The CLI measures its input, takes -padding and reports the format. Test data: seven format 2 fixtures, padding vectors checked against math/big, format 2 CBOR vectors, and the mutation corpus in both formats with the 22 cases of the third list of spec §64, built without randomness by sealing the fixtures again with their known keys and nonces. The format 1 fixtures are kept byte for byte and never regenerated; the differential corpus keeps its 1825 cases and adds a block per format 2 fixture. The spec copy loses its "to be implemented" markers, and the READMEs, CHANGELOG, traceability and testdata/README.md follow v0.9. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 week ago
- `fixtures/<name>.json`: for each `.dkc`, its `format`, its SHA-256, the
release that opens it, the hex of the prelude, PUBLIC_HEADER and
CONTROL_CBOR, the DateKey, `capsule_id`, `header_binding`,
`payload_identity` (I_PAYLOAD, a test secret), the visible stanzas of each
age file, the identities or `.dkk` that open it, `payload_length` (L), the
exact extension data, and the result of every step of §63. In format 2 it
adds `padding` (the code) and `padded_length` (P) and, for `time_and_key`,
`access_key_stanza` and `identity_stanzas`: the index, from 0, in
`inner_stanzas` of the stanza that the `.dkk` and each identity of
`identities`, in order, open. The other stanzas are dummies; the official
vectors are the only place where that is recorded (spec §39).
- `fixtures/<name>.dkk.json`: for each `.dkk`, its SHA-256, `credential_id`,
`capsule_id`, `access_type`, `access_material` (a test secret),
`capsule_digest`, extensions and the capsule it opens.

Powered by TurnKey Linux.