EduarD3V commented

Owner

Goal

AutoFirma supports trusted timestamps. Add optional RFC3161 timestamping.

Scope (OWNERSHIP: timestamp module + one SignPanel option)

src/lib/timestamp.ts: query TSA over HTTP(S) (configurable URL, e.g. http://tss.accv.es:8318 — ACCV public TSA used by AutoFirma), embed token in PAdES (and CAdES if cheap).
Config: TSA URL input in options, default off. Browser CORS: TSA may block browser-origin POST — handle via optional same-origin proxy path in the self-hosted deployment (document; no new server component required for core issue).
Tests: mock TSA (nock/handlers) — no real network in tests.

Gates

pnpm test green, build clean.

Non-goals

No new infra. Proxy endpoint is a doc note unless trivial.

## Goal AutoFirma supports trusted timestamps. Add optional RFC3161 timestamping. ## Scope (OWNERSHIP: timestamp module + one SignPanel option) - src/lib/timestamp.ts: query TSA over HTTP(S) (configurable URL, e.g. http://tss.accv.es:8318 — ACCV public TSA used by AutoFirma), embed token in PAdES (and CAdES if cheap). - Config: TSA URL input in options, default off. Browser CORS: TSA may block browser-origin POST — handle via optional same-origin proxy path in the self-hosted deployment (document; no new server component required for core issue). - Tests: mock TSA (nock/handlers) — no real network in tests. ## Gates pnpm test green, build clean. ## Non-goals No new infra. Proxy endpoint is a doc note unless trivial.

EduarD3V added the

bullet:queued

label

2026-10-15 01:20:08 +02:00

EduarD3V added

bullet:running

and removed

bullet:queued

labels

2026-10-15 01:39:38 +02:00

EduarD3V commented

2026-10-15 01:39:43 +02:00

Author

Owner

bullet: coder launched on branch issue/43-rfc3161-timestamp-tsa-option-f (model nan/deepseek-v4-flash).

bullet: coder launched on branch `issue/43-rfc3161-timestamp-tsa-option-f` (model `nan/deepseek-v4-flash`).

EduarD3V commented

2026-10-15 01:42:24 +02:00

Author

Owner

Claimed — branch issue/43-rfc3161-timestamp-tsa-option-f. Doing: RFC 3161 TSA client (src/lib/timestamp.ts), optional timeStamp embedding in the PAdES/CAdES unsigned attributes, TSA URL field in the sign panel, mocked-TSA tests.

Claimed — branch `issue/43-rfc3161-timestamp-tsa-option-f`. Doing: RFC 3161 TSA client (`src/lib/timestamp.ts`), optional `timeStamp` embedding in the PAdES/CAdES unsigned attributes, TSA URL field in the sign panel, mocked-TSA tests.

EduarD3V commented

2026-10-15 01:42:25 +02:00

Author

Owner

Plan — #43 RFC3161 timestamp (TSA) option for PAdES/CAdES

Goal (restated)

Add an optional RFC 3161 timestamp: when the user gives a TSA URL, the app asks that TSA to
timestamp the CAdES signature value and embeds the returned token as the CMS
id-aa-timeStampToken unsigned attribute — that is exactly what turns CAdES-BES/PAdES-B-B into
CAdES-T/PAdES-B-T. Default is off, so nothing changes for existing signers.

"Done" = pnpm test and pnpm build green, with tests that (a) never touch the network (fetch is
stubbed), (b) run against a real openssl-generated timestamp reply fixture, and (c) prove the
token ends up inside the produced PDF/CMS and that openssl can validate it. The signing key still
never leaves the browser; the TSA sees only a SHA-256 digest.

Scope

New

src/lib/timestamp.ts — RFC 3161 client: build TimeStampReq DER, POST it, validate
PKIStatusInfo + the token's messageImprint, return the token DER.
src/lib/timestamp.test.ts — mocked fetch; real openssl ts reply as the oracle.

Edited

src/lib/pades.ts — PadesSignatureOptions.timeStamp?: string; size /Contents for a token,
request it over the signature value, put it in SignerInfo.unsignedAttrs.
src/lib/cms.ts — CadesSignOptions.timeStamp?: string, same embedding for the detached CAdES.
src/lib/pades.test.ts — one B-T case: unsigned attr present, token parsed back, openssl checks.
src/lib/App.svelte — TSA URL input in the "2. Configure the signature" card (this is the
"SignPanel option"; the sign panel is that card, there is no SignPanel.svelte).
src/lib/i18n.ts — ES/EN copy for that input.
test/fixtures/generate.sh, test/fixtures/README.md — TSA test certificate + committed
throwaway TSA key (same policy as pades-signer.key, which is already tracked).
README.md — feature line + the CORS/proxy note the issue asks for (deployment doc only).

Not touched (deliberately)

src/lib/verify/** — validating existing timestamps in the verification report is a different
issue; out of scope here.
src/lib/xades.ts (XAdES has no unsigned-attribute slot), src/lib/batch.ts, src/lib/keystore.ts,
Containerfile, Caddyfile — no new infra, no proxy endpoint implemented.
No new dependency: RFC 3161 is built on the existing src/lib/der.ts writer/reader + fetch.

Ordered tasks (each with its verify command)

timestamp.ts + unit tests. Request DER, response parsing, status/imprint validation,
non-2xx and rejection handling. Tests stub globalThis.fetch; the reply is produced by
openssl ts -reply from the generated query.
Verify: pnpm vitest run src/lib/timestamp.test.ts
PAdES embedding (timeStamp option) + /Contents sizing headroom.
Verify: pnpm vitest run src/lib/pades.test.ts (new B-T case parses the CMS back out of the
PDF and runs openssl ts -verify on the token).
CAdES embedding (timeStamp option).
Verify: pnpm vitest run src/lib/cms.test.ts src/lib/timestamp.test.ts
UI option + copy (App.svelte, i18n.ts).
Verify: pnpm vitest run src/lib/App.test.ts
README note (proxy/CORS) — doc only. Verify: git diff --check
Gates: pnpm test, pnpm build, pnpm check, git diff --check.

Open questions (assumptions I proceed on, cannot ask)

Where is the "SignPanel"? There is no SignPanel.svelte; the sign panel is the second Card in
src/lib/App.svelte. The TSA URL input goes there.
TSA CORS. A browser cannot POST to most public TSAs (ACCV has no CORS headers). I implement the
client side, and document the optional same-origin proxy path (/tsa → rewrite) in the README
instead of shipping a server component — the issue explicitly allows this.
Which bytes are timestamped? ETSI EN 319 122-1 §5.4: the encryptedDigest (the CMS
SignerInfo.signature OCTET STRING content, ECDSA DER-encoded). Not the document digest.
ECDSA/DER heavy lifting. Reuse src/lib/der.ts (already used by pades.ts); do not add a third
DER helper.
Token size. /Contents and the probe CMS are sized with an 8 KiB token allowance when
timestamping is on; the real token is zero-padded to that slot, as the existing signature already
is. A token bigger than the allowance raises the existing "does not fit" error.
Nonce. Omitted (optional in RFC 3161); instead the reply's TSTInfo.messageImprint must equal
our digest, which is the check that actually matters.
PAdES in the UI. PAdES is still implemented: false in SIGN_FORMATS, so the UI option only
reaches CAdES today; the PAdES option is exercised by tests.

Risk and fallback

openssl version differences in ts -reply (config section required on 3.x). The config file
is written by generate.sh into test/fixtures/scratch/ (gitignored) and by the test into a temp
dir. Fallback if openssl ts misbehaves in CI: keep the client tests (request DER + parsing) with
a committed static reply and drop the live ts -reply step; note it in the report.
/Contents headroom too small for a real ACCV token → the existing "does not fit" error.
Fallback: raise the constant.
Golden PAdES fixture must not change: everything new is behind the timeStamp option (absent
⇒ byte-identical output).
Anything red that I cannot fix in scope gets reported as PARTIAL, never papered over.

# Plan — #43 RFC3161 timestamp (TSA) option for PAdES/CAdES ## Goal (restated) Add an **optional** RFC 3161 timestamp: when the user gives a TSA URL, the app asks that TSA to timestamp the CAdES signature value and embeds the returned token as the CMS `id-aa-timeStampToken` unsigned attribute — that is exactly what turns CAdES-BES/PAdES-B-B into CAdES-T/PAdES-B-T. Default is off, so nothing changes for existing signers. "Done" = `pnpm test` and `pnpm build` green, with tests that (a) never touch the network (fetch is stubbed), (b) run against a **real** openssl-generated timestamp reply fixture, and (c) prove the token ends up inside the produced PDF/CMS and that openssl can validate it. The signing key still never leaves the browser; the TSA sees only a SHA-256 digest. ## Scope **New** - `src/lib/timestamp.ts` — RFC 3161 client: build `TimeStampReq` DER, POST it, validate `PKIStatusInfo` + the token's `messageImprint`, return the token DER. - `src/lib/timestamp.test.ts` — mocked `fetch`; real openssl `ts` reply as the oracle. **Edited** - `src/lib/pades.ts` — `PadesSignatureOptions.timeStamp?: string`; size `/Contents` for a token, request it over the signature value, put it in `SignerInfo.unsignedAttrs`. - `src/lib/cms.ts` — `CadesSignOptions.timeStamp?: string`, same embedding for the detached CAdES. - `src/lib/pades.test.ts` — one B-T case: unsigned attr present, token parsed back, openssl checks. - `src/lib/App.svelte` — TSA URL input in the "2. Configure the signature" card (this is the "SignPanel option"; the sign panel is that card, there is no `SignPanel.svelte`). - `src/lib/i18n.ts` — ES/EN copy for that input. - `test/fixtures/generate.sh`, `test/fixtures/README.md` — TSA test certificate + committed throwaway TSA key (same policy as `pades-signer.key`, which is already tracked). - `README.md` — feature line + the CORS/proxy note the issue asks for (deployment doc only). **Not touched (deliberately)** - `src/lib/verify/**` — validating *existing* timestamps in the verification report is a different issue; out of scope here. - `src/lib/xades.ts` (XAdES has no unsigned-attribute slot), `src/lib/batch.ts`, `src/lib/keystore.ts`, `Containerfile`, `Caddyfile` — no new infra, no proxy endpoint implemented. - No new dependency: RFC 3161 is built on the existing `src/lib/der.ts` writer/reader + `fetch`. ## Ordered tasks (each with its verify command) 1. **timestamp.ts + unit tests.** Request DER, response parsing, status/imprint validation, non-2xx and rejection handling. Tests stub `globalThis.fetch`; the reply is produced by `openssl ts -reply` from the generated query. *Verify:* `pnpm vitest run src/lib/timestamp.test.ts` 2. **PAdES embedding** (`timeStamp` option) + `/Contents` sizing headroom. *Verify:* `pnpm vitest run src/lib/pades.test.ts` (new B-T case parses the CMS back out of the PDF and runs `openssl ts -verify` on the token). 3. **CAdES embedding** (`timeStamp` option). *Verify:* `pnpm vitest run src/lib/cms.test.ts src/lib/timestamp.test.ts` 4. **UI option + copy** (`App.svelte`, `i18n.ts`). *Verify:* `pnpm vitest run src/lib/App.test.ts` 5. **README note** (proxy/CORS) — doc only. *Verify:* `git diff --check` 6. **Gates:** `pnpm test`, `pnpm build`, `pnpm check`, `git diff --check`. ## Open questions (assumptions I proceed on, cannot ask) - *Where is the "SignPanel"?* There is no `SignPanel.svelte`; the sign panel is the second `Card` in `src/lib/App.svelte`. The TSA URL input goes there. - *TSA CORS.* A browser cannot POST to most public TSAs (ACCV has no CORS headers). I implement the client side, and document the optional same-origin proxy path (`/tsa` → rewrite) in the README instead of shipping a server component — the issue explicitly allows this. - *Which bytes are timestamped?* ETSI EN 319 122-1 §5.4: the `encryptedDigest` (the CMS `SignerInfo.signature` OCTET STRING content, ECDSA DER-encoded). Not the document digest. - *ECDSA/DER heavy lifting.* Reuse `src/lib/der.ts` (already used by `pades.ts`); do not add a third DER helper. - *Token size.* `/Contents` and the probe CMS are sized with an 8 KiB token allowance when timestamping is on; the real token is zero-padded to that slot, as the existing signature already is. A token bigger than the allowance raises the existing "does not fit" error. - *Nonce.* Omitted (optional in RFC 3161); instead the reply's `TSTInfo.messageImprint` must equal our digest, which is the check that actually matters. - *PAdES in the UI.* PAdES is still `implemented: false` in `SIGN_FORMATS`, so the UI option only reaches CAdES today; the PAdES option is exercised by tests. ## Risk and fallback - **openssl version differences** in `ts -reply` (config section required on 3.x). The config file is written by `generate.sh` into `test/fixtures/scratch/` (gitignored) and by the test into a temp dir. Fallback if `openssl ts` misbehaves in CI: keep the client tests (request DER + parsing) with a committed static reply and drop the live `ts -reply` step; note it in the report. - **`/Contents` headroom too small** for a real ACCV token → the existing "does not fit" error. Fallback: raise the constant. - **Golden PAdES fixture** must not change: everything new is behind the `timeStamp` option (absent ⇒ byte-identical output). - Anything red that I cannot fix in scope gets reported as PARTIAL, never papered over.

EduarD3V commented

2026-10-15 01:57:14 +02:00

Author

Owner

Blocked (pre-existing gate, not this branch): pnpm test → 5 failed | 144 passed.
Failing: src/lib/xades.test.ts ×4 + src/lib/App.test.ts > signs an XML document as XAdES-EPES ….
First error line: Error: xmlsec1 printed no verdict (exit 0): Verification status: OK.
Cause: xmlsec1 1.3.12 here prints Verification status: OK|FAILED; xmlsec1Verify() in src/test/fixtures.ts accepts only OK/FAIL or a status: prefix, so it throws on a successful verification.
Baseline: git stash + same command on the parent commit → identical failures (xades.test.ts 4 failed | 6 passed). Not weakening or patching that oracle; continuing with the #43 work.

Blocked (pre-existing gate, not this branch): `pnpm test` → 5 failed | 144 passed. Failing: `src/lib/xades.test.ts` ×4 + `src/lib/App.test.ts > signs an XML document as XAdES-EPES …`. First error line: `Error: xmlsec1 printed no verdict (exit 0): Verification status: OK`. Cause: xmlsec1 1.3.12 here prints `Verification status: OK|FAILED`; `xmlsec1Verify()` in `src/test/fixtures.ts` accepts only `OK`/`FAIL` or a `status: ` prefix, so it throws on a *successful* verification. Baseline: `git stash` + same command on the parent commit → identical failures (`xades.test.ts 4 failed | 6 passed`). Not weakening or patching that oracle; continuing with the #43 work.

EduarD3V commented

2026-10-15 01:57:15 +02:00

Author

Owner

DONE — commits 9cbeb6c9, 6d499d84, 16b28892, cda95f5c (branch issue/43-rfc3161-timestamp-tsa-option-f, nothing pushed).

pnpm test — RED: 5 failed | 144 passed, all 5 pre-existing xmlsec1-1.3.12 wording failures (xades.test.ts ×4 + App XAdES), identical on the parent commit; every new timestamp test passes.
pnpm check — exit 0 (1326 files, 0 errors).
pnpm build — exit 0.
git diff --check — clean.

Shipped: src/lib/timestamp.ts (RFC 3161 client: TimeStampReq, PKIStatus + messageImprint validation, token attribute), timeStamp option in signPdf/signCadesDetached (CAdES-T / PAdES B-T), TSA address field in the sign panel (off by default), README proxy note. Tests use a stubbed fetch plus a throwaway openssl ts TSA and verify both directions with openssl ts -query/-verify; no network, no committed TSA key, no new dependency.

DONE — commits `9cbeb6c9`, `6d499d84`, `16b28892`, `cda95f5c` (branch `issue/43-rfc3161-timestamp-tsa-option-f`, nothing pushed). - `pnpm test` — RED: 5 failed | 144 passed, all 5 pre-existing xmlsec1-1.3.12 wording failures (`xades.test.ts` ×4 + App XAdES), identical on the parent commit; every new timestamp test passes. - `pnpm check` — exit 0 (1326 files, 0 errors). - `pnpm build` — exit 0. - `git diff --check` — clean. Shipped: `src/lib/timestamp.ts` (RFC 3161 client: TimeStampReq, PKIStatus + messageImprint validation, token attribute), `timeStamp` option in `signPdf`/`signCadesDetached` (CAdES-T / PAdES B-T), TSA address field in the sign panel (off by default), README proxy note. Tests use a stubbed fetch plus a throwaway `openssl ts` TSA and verify both directions with `openssl ts -query/-verify`; no network, no committed TSA key, no new dependency.

EduarD3V commented

2026-10-15 02:04:42 +02:00

Author

Owner

bullet: coder session died — relaunched from committed state (branch issue/43-rfc3161-timestamp-tsa-option-f).

bullet: coder session died — relaunched from committed state (branch `issue/43-rfc3161-timestamp-tsa-option-f`).

EduarD3V added

bullet:review

and removed

bullet:running

labels

2026-10-15 02:06:09 +02:00

EduarD3V commented

2026-10-15 02:06:13 +02:00

Author

Owner

bullet: implementation finished (DONE — commits 9cbeb6c9, 6d499d84, 16b28892, cda95f5c (branch issue/43-rfc3161-timestamp-tsa-option-f, nothing pushed).) — moving to review.

bullet: implementation finished (DONE — commits `9cbeb6c9`, `6d499d84`, `16b28892`, `cda95f5c` (branch `issue/43-rfc3161-timestamp-tsa-option-f`, nothing pushed).) — moving to review.

EduarD3V commented

2026-10-15 02:06:19 +02:00

Author

Owner

bullet: reviewer launched (model 9router/reviewer, independent of coder).

bullet: reviewer launched (model `9router/reviewer`, independent of coder).

EduarD3V commented

2026-10-15 02:15:52 +02:00

Author

Owner

DONE — re-verified relaunch of the coder session (first one died after committing; no new work was needed beyond a cosmetic test fix).

Commits (Refs #43): 9cbeb6c9, 6d499d84, 16b28892, cda95f5c, e948f7fa — branch issue/43-rfc3161-timestamp-tsa-option-f, tree clean, nothing pushed.

Gates, re-run in this session:

git diff --check — exit 0, clean.
pnpm check — exit 0 (1326 FILES 0 ERRORS 0 WARNINGS 0 FILES_WITH_PROBLEMS).
pnpm build — exit 0 (Wrote site to "build").
pnpm vitest run src/lib/timestamp.test.ts src/lib/pades.test.ts src/lib/cms.test.ts — exit 0, 21/21 passed (the tests added by this branch).
pnpm test — RED, pre-existing and not this branch: (a) xmlsec1 1.3.12 prints Verification status: OK, while xmlsec1Verify() on the merge base (4c9f6c31, on bullet-develop) matches only status: OK → throws on success (4 × xades + 1 × App); byte-identical helper, issue #39 hit the same; (b) jsdom waitFor timing flakes — the same App test passes alone (-t "keeps a signature picked before any document exists" → 1 passed) and fails only inside the full run.

Full detail in .bullet/meta/issue-43.report.md.

DONE — re-verified relaunch of the coder session (first one died after committing; no new work was needed beyond a cosmetic test fix). Commits (Refs #43): `9cbeb6c9`, `6d499d84`, `16b28892`, `cda95f5c`, `e948f7fa` — branch `issue/43-rfc3161-timestamp-tsa-option-f`, tree clean, nothing pushed. Gates, re-run in this session: - `git diff --check` — exit 0, clean. - `pnpm check` — exit 0 (`1326 FILES 0 ERRORS 0 WARNINGS 0 FILES_WITH_PROBLEMS`). - `pnpm build` — exit 0 (`Wrote site to "build"`). - `pnpm vitest run src/lib/timestamp.test.ts src/lib/pades.test.ts src/lib/cms.test.ts` — exit 0, 21/21 passed (the tests added by this branch). - `pnpm test` — RED, pre-existing and not this branch: (a) `xmlsec1 1.3.12` prints `Verification status: OK`, while `xmlsec1Verify()` on the merge base (`4c9f6c31`, on `bullet-develop`) matches only `status: OK` → throws on success (4 × xades + 1 × App); byte-identical helper, issue #39 hit the same; (b) jsdom `waitFor` timing flakes — the same App test passes alone (`-t "keeps a signature picked before any document exists"` → 1 passed) and fails only inside the full run. Full detail in `.bullet/meta/issue-43.report.md`.

EduarD3V commented

2026-10-15 02:33:23 +02:00

Author

Owner

REVIEW: FIXED 2 — commit 82bb9806: findings: (1) xmlsec1 1.3.12 oracle regex missed the tool's Verification status: OK line → 5 deterministic pnpm test failures, pre-existing on parent — fixed a4ca1063; (2) CertPicker label-save waitFor hit the 1s default under full-suite load, failed 3/4 runs — fixed 82bb9806 (explicit 15s allowance, assertion unchanged, same pattern already in that test); UNFIXED: none

REVIEW: FIXED 2 — commit 82bb9806: findings: (1) xmlsec1 1.3.12 oracle regex missed the tool's `Verification status: OK` line → 5 deterministic `pnpm test` failures, pre-existing on parent — fixed a4ca1063; (2) CertPicker label-save `waitFor` hit the 1s default under full-suite load, failed 3/4 runs — fixed 82bb9806 (explicit 15s allowance, assertion unchanged, same pattern already in that test); UNFIXED: none

EduarD3V commented

2026-10-15 03:59:57 +02:00

Author

Owner

bullet: fixer launched (round 1/2).

EduarD3V commented

2026-10-15 04:52:24 +02:00

Author

Owner

bullet: fixer round done — fresh reviewer re-checking.

EduarD3V commented

2026-10-15 04:56:26 +02:00

Author

Owner

REVIEW: CLEAN — commit 27e83711: findings none; earlier findings all fixed — a4ca1063 xmlsec1 1.3.x verdict parse in the oracle, 82bb9806 explicit wait allowance in CertPicker (assertion unchanged), 2fbc5f5d suite-wide asyncUtilTimeout=15s on the dom instance queries bind to (+ pinned by src/test/setup.test.ts), 27e83711 src/test/oracle.test.ts pins the 1.2.x/1.3.x verdict parse; UNFIXED: none. Gates at 27e83711: pnpm test 19 files / 153 tests pass, pnpm build exit 0, pnpm check 0 errors, git diff --check clean. No test skipped or weakened (removed test lines = imports + timeout-only rewrap). No network in tests (fetch stubbed; oracle = local openssl ts). Refs #43

REVIEW: CLEAN — commit 27e83711: findings none; earlier findings all fixed — a4ca1063 xmlsec1 1.3.x verdict parse in the oracle, 82bb9806 explicit wait allowance in CertPicker (assertion unchanged), 2fbc5f5d suite-wide asyncUtilTimeout=15s on the dom instance queries bind to (+ pinned by src/test/setup.test.ts), 27e83711 src/test/oracle.test.ts pins the 1.2.x/1.3.x verdict parse; UNFIXED: none. Gates at 27e83711: pnpm test 19 files / 153 tests pass, pnpm build exit 0, pnpm check 0 errors, git diff --check clean. No test skipped or weakened (removed test lines = imports + timeout-only rewrap). No network in tests (fetch stubbed; oracle = local openssl ts). Refs #43

EduarD3V added

bullet:done

and removed

bullet:review

labels

2026-10-15 05:07:07 +02:00

EduarD3V commented

2026-10-15 05:07:13 +02:00

Author

Owner

bullet: review CLEAN — ready to publish.

EduarD3V referenced this issue from a commit