provared 0.4.0 → 0.6.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/docs/guide.md CHANGED
@@ -6,7 +6,7 @@ Everything about the format, the library, the checker, the stub writer beside an
6
6
 
7
7
  ## Status
8
8
 
9
- **Version 0.4.0, a draft.** Published as it is, by one author in their
9
+ **Version 0.6.0, a draft.** Published as it is, by one author in their
10
10
  own time; fixes as time allows. The format may still change, so do not
11
11
  rely on a draft record staying checkable.
12
12
 
@@ -21,7 +21,7 @@ An Internet-Draft is a working document, not a standard.
21
21
  |---|---|---|
22
22
  | **Slip** | A short permission a person signs with a passkey: which agent, which actions, what limits and conditions, with whom, from when until when, and why | Built |
23
23
  | **Stub** | A signed receipt for each action, pointing back to its slip, countersigned by the other side where possible, and approved by the person where the slip asks for it | Built |
24
- | **Seal** | The stubs chained together and gathered under one top fingerprint, signed by whoever keeps the book and time-stamped by an outside service | Built. A public place where seals are kept is not |
24
+ | **Seal** | The stubs chained together and gathered under one top fingerprint, signed by whoever keeps the book, time-stamped by an outside service, and signed for by witnesses | Built, with the part of a witness that decides and signs. A public place where seals are kept is not |
25
25
  | **Show** | One page shown to a checker with proof that it is in the book, and when the book was sealed, with the names it does not need kept covered | Built. Only a slip's names and purpose can be covered so far |
26
26
 
27
27
  ## What is here
@@ -151,9 +151,9 @@ be loaded from anywhere. The tool prints the file's SHA-256 fingerprint,
151
151
  so that a copy can be compared with it. (Tried on 8 October 2026 in the
152
152
  three browser builds this repository is tested with, and on 13 October
153
153
  2026 in Chrome 155, Edge 155 and Brave 155.) The file for
154
- version 0.4.0 is attached to the release on GitHub. Its SHA-256
154
+ version 0.6.0 is attached to the release on GitHub. Its SHA-256
155
155
  fingerprint is
156
- `49da6617ff493ddda2b2b65bfb6bedebdf8cf07ea6faa32b039f065eda207307`.
156
+ `ae48c425616d70849aa8d43b1d71ef7ee56736da3ba6a00eaa4375c428e60a25`.
157
157
 
158
158
  **Run the demonstration in a browser.**
159
159
 
@@ -168,8 +168,9 @@ is invented: nothing is bought, sent or deleted. The checking
168
168
  page on its own is at `http://localhost:8787/check`. It takes a record
169
169
  file, the fingerprints you already trust (keys, time-stamp services,
170
170
  blocks), and what you were handed with the record: the person's own
171
- copies of cancellations, and disclosures for covered fields. All of it is
172
- read in the browser and sent nowhere.
171
+ copies of cancellations, seals of the book held elsewhere, and
172
+ disclosures for covered fields. All of it is read in the browser and
173
+ sent nowhere.
173
174
 
174
175
  ## What a slip can say
175
176
 
@@ -236,6 +237,214 @@ The fingerprints for the sample are in
236
237
  `samples/office-supplies.expected.json`. The sample's time-stamp service
237
238
  is made up, as its passkey is.
238
239
 
240
+ ### Seals held elsewhere
241
+
242
+ Whoever keeps the book can cut off its end, with the last seal, and what
243
+ is left still checks. Anyone who keeps a copy of a seal elsewhere (the
244
+ person, an auditor, a second copy out of the recorder's reach) can hand
245
+ it to the checker beside the book, as the entry it is in the book:
246
+
247
+ ```
248
+ provared-check record.jsonl --seal seal.json --sealer <fingerprint> --issuer <thumbprint>
249
+ ```
250
+
251
+ In the library: `seals`, up to 64. A seal at entry n covers the n
252
+ entries before it, so the book must hold that very seal at entry n. A
253
+ book that ends before it is reported as cut short (`book-cut-short`); a
254
+ book that holds another line there, as changed (`seal-not-in-book`).
255
+ Both are problems with the book. A seal handed over is checked as any
256
+ seal is, and counts only where you named its recorder (`--sealer`, or
257
+ `sealKeys`); one that does not pass its own check is a problem with the
258
+ whole check. On a device without SLH-DSA (the Python version, and
259
+ browsers that have ML-DSA, such as Edge and Brave 155 when tested) it
260
+ counts on its Ed25519 and ML-DSA-87 signatures, and says so; on a device
261
+ without ML-DSA it is not used. The seal is compared with the book by its
262
+ content: the same content with other signatures is the same seal, and
263
+ the answer says that the signatures differ. Nothing in a seal names its
264
+ book, so hand over only seals of the book being checked: a seal of
265
+ another book by the same recorder (for example the book it kept before
266
+ it started a new one) is reported as `seal-not-in-book` or
267
+ `book-cut-short`. Without `--sealer`, the seals handed over are not
268
+ compared with the book, and the command-line checker says so on a line
269
+ of its own and gives no pass.
270
+ Its own time-stamps are shown but are not used for the entries of the
271
+ book. The answer lists each one in `held`, after any cancellations. A
272
+ Show does not use one. What it does not catch: a book cut back after the
273
+ latest seal anyone holds, or a book never shown to anyone who holds one.
274
+
275
+ ### Witnesses
276
+
277
+ A witness is a party, with a key set of its own, that keeps for each
278
+ book the latest seal it has signed for, and signs a short statement for
279
+ a new seal only where the new seal extends that one (format
280
+ description, section 29). The recorder sends it the seal and a
281
+ consistency proof (RFC 9162, section 2.1.4); the statement goes beside
282
+ the seal on its line, in `witnesses`, up to four, or is handed to the
283
+ checker later with a copy of the seal (`--seal` with a file of
284
+ `{"seal": ..., "witnesses": [...]}`). Name the witnesses you trust by
285
+ the fingerprints of their key sets:
286
+
287
+ ```
288
+ provared-check record.jsonl --witness <fingerprint> --sealer <fingerprint> --issuer <thumbprint>
289
+ ```
290
+
291
+ In the library: `witnesses`. The answer's `summary.witnessed` gives the
292
+ latest seal for which a statement from a witness you trust counts, the
293
+ entries it covers, those witnesses, and every seal witnessed; the
294
+ command-line checker prints it as "Witnessed by a witness you trust?".
295
+ A statement that does not check, or that speaks of another seal, is a
296
+ problem (`witness-invalid`). A statement counts only where you also
297
+ named the passkey you trust (`--issuer`) and the recorder you expect
298
+ (`--sealer`), the book starts with a permission signed by that passkey,
299
+ and every permission before the seal is from that passkey for the same
300
+ agent (a Show must hold every page the seal covers): a witness knows a book only by the passkey and the agent of
301
+ the permission it starts from and its recorder's keys, and the recorder
302
+ chooses the first line and its keys. Being witnessed excuses nothing
303
+ and is not needed for a pass. What it adds: a second history of the
304
+ entries before a witnessed seal, sealed by the same recorder and
305
+ starting from any permission that passkey gave that agent (a renewal
306
+ included), shown to someone else, would need that witness to sign for it
307
+ too. What it does not: a history sealed with other keys, even keys you
308
+ also named, or one that starts from a permission for another agent;
309
+ anything never shown to the witness; or the time. To see whether the end
310
+ of the book is missing, ask the witness for the latest seal it holds for
311
+ this recorder, passkey and agent, and hand that seal over beside the book
312
+ it is of. One recorder key set keeps one sequence for each person and
313
+ agent: books kept side by side for one agent need recorder keys of
314
+ their own.
315
+
316
+ A recorder may keep the record of one permission in several books: a
317
+ seal says which in `book` (2, 3 and so on; the first book has no
318
+ number), and the first seal of a later book names the last seal of the
319
+ book before. `writeSeal` takes `book`. A witness treats every book of one
320
+ recorder, person and agent as one sequence: it signs for a new book only
321
+ where its first seal starts from the latest seal it holds, and never
322
+ again for the book before. A checker of one book does not see the book
323
+ before; handed that book's last seal, it says whether this book follows
324
+ on from it (format description, section 21.6). A seal handed over that
325
+ gives another book number than the book is not compared with it, and
326
+ gives no pass: the command-line checker says "Seals held elsewhere
327
+ compared? NO" and does not end with code 0, even for an honest book.
328
+ To check book n − 1 against the seal book n starts from, check book
329
+ n − 1 with that seal handed over beside it.
330
+
331
+ The library holds a witness: `openWitness({ by, signers, state })`, with
332
+ `witness(entry, proof, firstLine)`, which signs only for a seal whose
333
+ three signatures it checked and that extends the seal it holds, and
334
+ hands back its state with each statement, to be saved before the
335
+ statement is given out; `update`, `ready`, `latest` and `state`.
336
+ `witnessProof(book, size, from)` makes the proof to send it (`from`:
337
+ the size of the seal the witness holds of the same book, or none for a
338
+ first seal and for the first seal of the next book), and `bookIdentity`
339
+ names a sequence as a witness knows it: by the recorder's keys with the
340
+ passkey and the agent of the permission on the book's first line. Whoever can change a witness's
341
+ state can make it forget a sequence: keep it as safely as its keys. A
342
+ witness needs SLH-DSA, so it needs Node.js; the Python version has none.
343
+
344
+ ### A central sealer
345
+
346
+ The recorder's three keys need not sit on the agent's computer. A
347
+ central sealer, run by a team other than the one that runs the agent,
348
+ holds them and seals books it never sees: the recorder sends it the
349
+ leaf fingerprint of each new line (`leafFingerprint(line)`: the SHA-256
350
+ of the byte 0x00 and the line), and asks for a seal; the sealer builds
351
+ the same tree as the checker, signs a seal over its own number of
352
+ entries and top fingerprint, and hands back the seal's entry, which the
353
+ recorder adds to its book with the stub writer's `add` (format
354
+ description, section 21.7). A sealer sent a wrong fingerprint signs a
355
+ seal the book refuses (`seal-mismatch`).
356
+
357
+ ```js
358
+ import { leafFingerprint, openSealer, toBase64url } from 'provared';
359
+
360
+ const sealer = await openSealer({ by: { keys, name: 'Our central sealer' }, signers, state });
361
+ const first = toBase64url(await leafFingerprint(lines[0])); // a book is known by its first line
362
+ await sealer.extend(first, 0, await Promise.all(lines.map(leafFingerprint)));
363
+ const made = await sealer.seal(first, { previous: null }); // the seal the book's latest seal names, or null
364
+ // save made.state first, then hand made.entry to the book
365
+ ```
366
+
367
+ `extend(book, at, leaves)` takes lines only at the place it holds, so a
368
+ line is never taken twice. `seal(book, { previous })` signs only where
369
+ `previous` is the latest seal it holds of that book: a sealer whose
370
+ state is behind the book signs nothing until it is brought up to date
371
+ with `update(book, entry, leaves)` (the book's latest seal, and the leaf
372
+ fingerprints since the latest seal it holds); a book that lost a seal
373
+ the sealer gave takes it back with `latest(book)`. A later book is named
374
+ `{ first, number }`, and its first `extend` names the book it follows
375
+ (`{ follows }`); its first seal names the last seal of that book, which
376
+ is then closed, and so is a book brought up to date from such a first
377
+ seal (bring books up to date in their order: a later book's first seal
378
+ is refused, `not-up-to-date`, while the sealer holds the book before at
379
+ an earlier seal than the one it names). Every book with one first line
380
+ is one chain: a book whose first line the sealer holds starts only as
381
+ the next book after the latest with that first line (`not-next-book`). A sealer that serves several
382
+ recorders gives each its own calls, `sealer.forRecorder(name)`: a book,
383
+ and every book after it, belongs to the recorder that started it, and
384
+ another recorder's calls about it are refused (`other-recorder`) or
385
+ answered as if the sealer did not hold it. Recorders that do not trust
386
+ each other are given rules, `sealer.forRecorder(name, { agents, books })`:
387
+ the fingerprints of the key sets of the agents whose permissions each may
388
+ start books with (an agent belongs to one recorder), and the most books
389
+ each may hold. Then every book is started with its first line itself
390
+ (`extend(book, 0, leaves, { line })`, or `update(book, entry, leaves, {
391
+ line })` for a book brought in), which must be a permission for one of
392
+ the recorder's own agents (`agent-not-allowed`, or `other-recorder` for
393
+ another recorder's agent); so no recorder can start a book whose first
394
+ line is a permission for another recorder's agent. `releaseUnsealed()` takes off the
395
+ books that hold no seal, for which nothing was signed. Save the state
396
+ handed back with a seal, with a book started (`extend`) and with
397
+ `update` before the answer is given out, and the one `releaseUnsealed`
398
+ hands back. Of a seal entry brought in with `update`, the sealer keeps
399
+ the seal, whose three signatures it checks, and its time-stamps, read
400
+ for their form only: it checks no time-stamp's signature, so one that
401
+ `latest` gives back is not vouched for by the sealer. A book is
402
+ kept in the state at its latest seal, or with none at its first line:
403
+ started again, the sealer asks for the lines since. It keeps at most
404
+ 1,000 books, about 52 KB of state for each sealed book. It checks every
405
+ seal it signs, so it needs Node.js; the Python version has
406
+ `leaf_fingerprint` only.
407
+
408
+ What a central sealer adds, and what still needs a witness. With its
409
+ state kept, it never signs two different histories of one first line.
410
+ With rules, it starts a recorder's books only from permissions for that
411
+ recorder's own agents, and so learns each first line. It sees only
412
+ fingerprints after the first line, so it cannot tell that a book which
413
+ starts from another first line (another permission for the same agent,
414
+ or, with no rules, any record the agent's side can sign) repeats lines
415
+ of a book it has sealed. A sealer that loses its state, or is put back
416
+ to an older copy, can be made to sign a second history, and two states
417
+ kept with one key set are two sealers. A witness is the check for these
418
+ (format description, sections 21.7 and 29).
419
+
420
+ One case a witness does not settle. A recorder, through its own calls
421
+ alone, can start a book from a permission for one of its own agents and
422
+ copy into it the permissions and stubs of another recorder's agent,
423
+ leaving out any it chooses (a stub over a limit, say); the sealer seals
424
+ it. Where one person's passkey signed the permissions for both agents,
425
+ a reader who names that passkey and the sealer's key set finds the book
426
+ intact and within its slips. A witness statement on its seal is not
427
+ counted (the seal covers a permission for an agent other than the first
428
+ line's), so the seal reads as not witnessed, but nothing reports why.
429
+ What shows it: a seal of the other recorder's own book, held elsewhere
430
+ and handed to the checker with this book (`seals`, `--seal`), gives
431
+ `book-cut-short` or `seal-not-in-book`; or, where the two books'
432
+ numbers differ, it cannot be compared, and the checker says so and
433
+ gives no pass. So readers of books from a
434
+ sealer shared by recorders that do not trust each other should hold
435
+ each recorder's own seals elsewhere, or the organisation should run a
436
+ sealer, with its own key set, for each recorder (format description,
437
+ section 21.7).
438
+
439
+ The folder `sealer/` of the repository holds `provared-sealer`, a small
440
+ program around it: it listens on 127.0.0.1, takes questions only from the
441
+ recorders it was given (each proves that it holds its token, which never
442
+ travels), names each recorder with its agents and its number of books,
443
+ keeps one state for all of them, keeps its keys in a file for its owner
444
+ alone or reaches them through signer modules named in its settings, and
445
+ makes sure the state is on the disk before it answers a question that
446
+ changed it. See its README.
447
+
239
448
  ### Knowing that a block is part of the chain
240
449
 
241
450
  **Experimental.** `provared-headers` is new in version 0.1.0 and may
@@ -533,7 +742,8 @@ A call that the slip does not allow throws `NotTaken`, with the answer
533
742
  of the check before acting, and the tool is not run. Where the slip asks
534
743
  for the person's own approval, a tool's `approve` function is asked for
535
744
  it once. By default the stub of a call names one document,
536
- `arguments`: the fingerprint of the call's arguments, written in one
745
+ `provared.arguments` (the reserved name of section 27 of the format
746
+ description): the fingerprint of the call's arguments, written in one
537
747
  form (RFC 8785). So the record fixes exactly what was asked for without
538
748
  holding it; `argumentsFingerprint` works the same fingerprint out again.
539
749
  The arguments must be JSON: a list with a gap in it, or with a named
@@ -551,6 +761,132 @@ not a promise. The numbers `0` and `-0` share a fingerprint, as RFC
551
761
  though the tool has run and its stub is in the book. `recordTools`
552
762
  names no agent software, and works with any function.
553
763
 
764
+ ### Denials: the calls the agent's side refused
765
+
766
+ A book shows what the agent did, not how often it was stopped. Opened
767
+ with `denials: true`, the stub writer also writes a **denial** whenever
768
+ `act` does not take an action: the agent's side's own signed statement
769
+ that it did not take it, and why (format description, section 28).
770
+
771
+ ```js
772
+ const recorder = await openRecorder({ book, slip, privateKeys, issuerKeys, denials: true });
773
+
774
+ const outcome = await recorder.act({ action: 'provared.data.delete' }, async () => removeTheFiles());
775
+ outcome.done; // false: the slip does not allow it
776
+ outcome.denial.written; // true: the book now holds a denial, by "check", with the reason "action-not-allowed"
777
+
778
+ // A refusal that came from elsewhere: the person said no, or a rule of the software around the agent did.
779
+ await recorder.deny({ action: 'provared.order.place', amount: { unit: 'GBP', value: 900 } }, { source: 'person' });
780
+ ```
781
+
782
+ A denial names what was asked for (its action, amount and service) and
783
+ who said no: `check` (the check before acting, with the codes it gave),
784
+ `person` or `rule`. Its documents are fingerprints only; arguments are
785
+ never written into it. `more.denial: false` asks `act` to write none
786
+ for one call. A writer writes at most `denialsPerHour` (600 by default)
787
+ in any hour by its clock. Each denial it writes, or is handed through
788
+ `add`, counts at the clock's time when it joins the book, whatever date
789
+ it states; those of its own chain that the book already held when it
790
+ was opened count at the dates they state. Past that it writes none, and
791
+ each answer's `notWritten` says how many it left out since it was
792
+ opened. A writer opened again on a book counts the book's denials at
793
+ their stated dates, so a caller who dates denials back and then opens
794
+ the writer again can get a second burst (5 and then 5 more within 3,302
795
+ seconds, at a limit of 5). At 600 an hour, an agent that is refused again and again still
796
+ fills a book's 100,000 entries in about a week: the limit slows that,
797
+ it does not stop it. A stub is never kept out by that limit, and
798
+ nothing that goes wrong in writing a denial makes `act` or `deny` fail:
799
+ the answer says what was done. `deny` does throw where the call itself
800
+ is malformed: a request that is not an object, or that cannot be copied
801
+ as plain data, a `when` that is neither
802
+ a number nor a date, `codes` that cannot be copied, or a call from
803
+ inside an action, which is refused as every call from there is. It
804
+ writes nothing for a writer opened without `denials: true`, and takes a
805
+ `when` within 300 seconds of the clock. It records what its caller
806
+ states: it does not check that the action was not taken, so a caller
807
+ can write a denial for a request that also has a stub.
808
+
809
+ `recordTools` passes it on: with a writer that writes denials, a call
810
+ that is not run leaves one, and `NotTaken` holds what was done
811
+ (`denial`). Where a tool can ask the person for approval, no denial is
812
+ written until it is known that the call will not run: so one call
813
+ leaves a stub or a denial, never both. A checker shows each denial as
814
+ the agent's side's own word, counts it (`summary.counts.denials`), and
815
+ never counts it against the agent. An agent can simply not write one,
816
+ so a book without denials shows nothing about refusals.
817
+
818
+ ### Binding a stub to the platform's own record
819
+
820
+ Where the other side will not countersign, a stub can still be matched
821
+ with the other system's own log of the call: name, among its documents,
822
+ the fingerprint of an identifier that both records hold. Document names
823
+ that begin `provared.` are reserved for this (format description,
824
+ section 27); any other such name is refused.
825
+
826
+ | Name | The fingerprint of | Made with |
827
+ |---|---|---|
828
+ | `provared.arguments` | The call's arguments, in the one form of RFC 8785 | `argumentsFingerprint` |
829
+ | `provared.http-request` | The request: its method, its address exactly as sent, and its body where it has one; no header | `httpRequestFingerprint` |
830
+ | `provared.http-response` | The answer: its status code, and its body where it has one; no header | `httpResponseFingerprint` |
831
+ | `provared.trace-id` | The trace-id of the W3C Trace Context `traceparent` header the agent's side sent (its form is not checked) | `identifierFingerprint` |
832
+ | `provared.request-id` | An identifier the agent's side chose and sent with the call | `identifierFingerprint` |
833
+ | `provared.event-id` | An identifier the platform gave for the call | `identifierFingerprint` |
834
+ | `provared.error` | The error text the agent's software was given | `identifierFingerprint` |
835
+
836
+ `DOCUMENT_NAMES` lists them, each with the exact text it stands for. An
837
+ HTTP request or answer is written as the signature base of RFC 9421
838
+ (nothing is signed with it): the address is taken exactly as it was
839
+ sent, never rewritten, so give it in the form that was sent. A body is
840
+ the bytes RFC 9530 digests: after any content coding (such as gzip),
841
+ before any transfer coding; a body of `null` is no body. The three
842
+ identifiers may be named more than once in a stub; every other reserved
843
+ name once at most. A trace-id is fingerprinted as given: its form (32
844
+ lower-case hexadecimal characters in the W3C standard) is not checked.
845
+
846
+ An identifier that the platform gives only in its answer cannot be in a
847
+ stub signed before the action. `act` takes `detailsAfter` for it:
848
+
849
+ ```js
850
+ import { identifierFingerprint } from 'provared';
851
+
852
+ const outcome = await recorder.act(
853
+ { action: 'provared.data.change', with: 'tickets' },
854
+ async () => updateTheTicket(), // gives the platform's answer
855
+ { detailsAfter: async (answer) => [{ name: 'provared.event-id', sha256: await identifierFingerprint(answer.requestId) }] },
856
+ );
857
+ outcome.detailsAfter; // { added: 1, problem: null }
858
+ ```
859
+
860
+ The stub is still written and checked with the whole book before the
861
+ action, so it is known to fit. After the action it is signed again with
862
+ up to four further documents, within the same call, so that no other
863
+ call comes between; it keeps the time the action began. If
864
+ `detailsAfter` fails, gives no answer in the time the other side is
865
+ given to countersign (`countersignWithin`), gives more than four
866
+ documents, or gives one that would be refused, the stub is written
867
+ without them, and `outcome.detailsAfter.problem` says why: an action
868
+ that was taken always gets its stub. The request may then name at most
869
+ 28 documents of its own, and no approval may be given, because an
870
+ approval covers exactly the documents the person saw. `detailsAfter`
871
+ must not call the writer.
872
+
873
+ A matching identifier shows that the two records speak of one call, not
874
+ that the call did what either says: the platform's identifier is the
875
+ platform's word. It binds the two only where the platform writes the
876
+ same identifier into its own log; which one a platform keeps is for you
877
+ to find out, so record both one you sent and one you were given where
878
+ you can. Only an identifier picks out one call: a fingerprint of a
879
+ request, an answer, the arguments or an error shows only that the
880
+ content matches, since identical requests, and identical answers (many a
881
+ `200` with the body `{}`), share one fingerprint. A platform that logs
882
+ the decoded body of a compressed message will not match the body's
883
+ fingerprint. An identifier that is short or easy to guess can be guessed
884
+ from its fingerprint.
885
+
886
+ `recordTools` names the arguments `provared.arguments` by default (before
887
+ version 0.6.0, `arguments`; every checker reads both); a tool can name
888
+ them otherwise through its own `details` function.
889
+
554
890
  ## How signing works
555
891
 
556
892
  - **The person** signs a slip with a passkey. A passkey signs a short
@@ -575,6 +911,107 @@ names no agent software, and works with any function.
575
911
  signatures were not checked, and does not show a pass. Node.js checks
576
912
  all of them.
577
913
 
914
+ ### Keys held elsewhere: signers
915
+
916
+ The keys of an agent, a service, an organisation or a recorder need not
917
+ sit in the memory of the program that writes the records. Wherever a
918
+ writer takes `privateKeys` (`openRecorder`, `writeStub`, `countersign`,
919
+ `writeRefusal`, `writeTerms`, `writePass`, `writeAcknowledgement`,
920
+ `writeVouching`, `writeWithdrawal` and `writeSeal`), each item may be a
921
+ key, as before, or a **signer**: an object with two members.
922
+
923
+ - `key`: the public key, as a JSON Web Key in one of the shapes the
924
+ format allows (format description, section 3.8).
925
+ - `sign(data)`: a promise of the signature over exactly those bytes, as a
926
+ `Uint8Array` or an `ArrayBuffer`.
927
+
928
+ A signer lets a key live in a key service or a hardware module. Keys and
929
+ signers may be mixed: an agent may keep its ML-DSA-87 key in a key
930
+ service and its Ed25519 key elsewhere. `signerFromKey(alg, privateKey,
931
+ key)` wraps a key held in memory as a signer, so that both take one path.
932
+
933
+ ```js
934
+ import { openRecorder, signerFromKey } from 'provared';
935
+
936
+ // "keyService.sign" stands for the service's own call. It must sign the
937
+ // bytes themselves, not a fingerprint of them.
938
+ const held = {
939
+ key: agentKeys[1], // the agent's ML-DSA-87 public key, as the slip names it
940
+ sign: (data) => keyService.sign('agent-ml-dsa-87', data),
941
+ };
942
+ const recorder = await openRecorder({
943
+ book, slip, issuerKeys,
944
+ privateKeys: [signerFromKey('Ed25519', ed25519PrivateKey, agentKeys[0]), held],
945
+ signWithin: 10000, // milliseconds for each signature; 30,000 by default
946
+ });
947
+ ```
948
+
949
+ A record made through a signer is the same record as one made with the
950
+ key itself: the same content, the same labels and the same Ed25519
951
+ signature. ML-DSA-87 and SLH-DSA signatures are made afresh each time,
952
+ with a key or through a signer, so those differ between any two
953
+ signings, as they always did.
954
+
955
+ A signer is not taken on trust:
956
+
957
+ - Its two members are read once. Its key must be exactly a key of the
958
+ method of its place in the list, as plain data.
959
+ - A signer for a method this device does not have built in is refused
960
+ before any signer is asked to sign, since its signatures could not be
961
+ checked here.
962
+ - Where the writer knows the key the record must be signed with, the
963
+ signer's key must be that key: the agent's keys as the slip names them,
964
+ or a helper's as its pass names them, in the stub writer; the keys
965
+ `by` names in `writeSeal`, `writeRefusal`, `writeTerms` and
966
+ `writeVouching`. A signer for another key is refused (`bad-key`)
967
+ before any signer is asked to sign. This catches the most likely
968
+ fault of a key held elsewhere: a mistyped key name in its settings.
969
+ The stub writer compares when it is opened, with the slip, or the
970
+ pass it acts under, as its book holds it. Opened with a pass that its
971
+ book does not yet hold, it compares nothing then, and refuses each
972
+ action (`pass-missing`), as it always did.
973
+ - `writeStub`, `countersign`, `writePass`, `writeAcknowledgement` and
974
+ `writeWithdrawal`, used by themselves, cannot know which key the
975
+ record must be signed with. There a signer for another key gives a
976
+ record that fails its own check, as a wrong key in memory always did.
977
+ - Every signature it gives is checked against its key before it is
978
+ used. A signature of the wrong length, one that does not check
979
+ (another key used, or other bytes signed, such as a fingerprint of
980
+ them), something other than bytes, or a `sign` that throws: no record
981
+ is made (`signature-invalid`), nothing is written to a book, and the
982
+ error names the place in the list, the method and the key's
983
+ thumbprint.
984
+ - In the stub writer, a signer is given `signWithin` milliseconds for
985
+ each signature (30,000 by default). After that the action is not
986
+ taken, nothing is written, the writer goes on to its next call, and a
987
+ late answer is not used. A signer that calls the writer is refused at
988
+ once, as an action that does so is. The other writers set no time of
989
+ their own: a signer that never answers leaves the call waiting, so give
990
+ the signer's own connection a time limit.
991
+
992
+ Where no record is made, signers earlier in the list may already have
993
+ signed, and a signer that ran out of time may still sign later. So a key
994
+ service's own log may hold signatures that match no record.
995
+
996
+ A key given as before is used exactly as before: a private Web Crypto
997
+ `CryptoKey` of the method of its place and, in a seal's SLH-DSA place, a
998
+ private Node.js `KeyObject`, or what Node.js reads as one (PEM text, DER
999
+ bytes with their format, or a JSON Web Key with its format). Anything
1000
+ else that is not a signer (a public key, a key of another method or
1001
+ kind, a value that only looks like a key) is refused (`bad-field`).
1002
+
1003
+ What a signer does not change:
1004
+
1005
+ - **The library cannot know how a key held elsewhere is kept**, who else
1006
+ can ask for signatures with it, or whether the service keeps it safe.
1007
+ It checks only that each signature fits the key the signer names.
1008
+ - **A record does not say where a key was kept.** A record signed
1009
+ through a key service reads the same as one signed with a key in
1010
+ memory, and a key set split between two places reads as one agent.
1011
+ - **Each signature costs one more check.** Measured on one computer on
1012
+ 15 October 2026: about 0.3 thousandths of a second for Ed25519, 0.6
1013
+ for ML-DSA-87 and 2 for SLH-DSA-SHA2-256s.
1014
+
578
1015
  ## Using the library
579
1016
 
580
1017
  ```js
@@ -626,15 +1063,15 @@ answer with a named reason. It keeps two things apart:
626
1063
  |---|---|
627
1064
  | `size`, `root` | How many entries the book holds, and its top fingerprint |
628
1065
  | `problems` | Problems with the book as a whole |
629
- | `entries` | One for each line: `index`, `kind` (`slip`, `stub`, `refusal`, `terms`, `seal`, `cancellation`, `acknowledgement`, `vouching`, `withdrawal`, `pass` or `unreadable`), `fingerprint`, `verified` (`all`, `some` or `none` of its signatures confirmed here), `problems`, `breaches`, `notes`, the signature states, and `content`. Where a time-stamp you trust on a later seal covers the entry: `existedBy`, the time by which it is shown to have existed. On a seal or a cancellation that passed its check: `stampedAt`, the time its own time-stamps give it |
630
- | `held` | One for each cancellation handed over beside the book: whether it passed its check and was used, whether the book holds it (`inBook`), the time its time-stamp gives, and the acknowledgements handed over with it (`acknowledgements`: who acknowledged, and when by their own word) |
1066
+ | `entries` | One for each line: `index`, `kind` (`slip`, `stub`, `refusal`, `terms`, `seal`, `cancellation`, `acknowledgement`, `vouching`, `withdrawal`, `pass`, `denial` or `unreadable`), `fingerprint`, `verified` (`all`, `some` or `none` of its signatures confirmed here), `problems`, `breaches`, `notes`, the signature states, and `content`. Where a time-stamp you trust on a later seal covers the entry: `existedBy`, the time by which it is shown to have existed. On a seal or a cancellation that passed its check: `stampedAt`, the time its own time-stamps give it |
1067
+ | `held` | One for each cancellation handed over beside the book (`kind` `cancellation`): whether it passed its check and was used, whether the book holds it (`inBook`), the time its time-stamp gives, and the acknowledgements handed over with it (`acknowledgements`: who acknowledged, and when by their own word). Then one for each seal held elsewhere (`kind` `seal`): whether it passed its check and was used, the entry it stands at (`place`), whether the book holds it there (`inBook`), its recorder's key set (`sealer`), its signatures and time-stamps |
631
1068
  | `summary.problemFound` | A problem was found somewhere in the record |
632
1069
  | `summary.fullyChecked` | Every signature was checked on this device |
633
1070
  | `summary.intact` | No problem was found **and** every signature was checked. The only answer that is a pass |
634
1071
  | `summary.methodsMissing` | The signing methods this device lacks, if any, for example `ML-DSA-87` |
635
1072
  | `summary.withinSlips` | Every stub was compared with its slip, and none is outside it |
636
1073
  | `summary.firstBreach` | The first entry that shows the agent outside its slip, if any |
637
- | `summary.counts` | Slips, stubs, countersigned stubs (the countersignature confirmed on this device), one-sided stubs (no countersignature; a stub that was not checked, because its slip failed, is neither), stubs the person approved, refusals, sets of terms, and the other kinds of entry |
1074
+ | `summary.counts` | Slips, stubs, countersigned stubs (the countersignature confirmed on this device), one-sided stubs (no countersignature; a stub that was not checked, because its slip failed, is neither), stubs the person approved, refusals, sets of terms, denials (the agent's side's own word, never counted against it), and the other kinds of entry |
638
1075
  | `summary.sealed` | How far the last time-stamp you trust reaches: how many entries, the time they existed by, and whether a service or a block (`by`) gave it. Otherwise nothing |
639
1076
  | `summary.limits` | What no check can show, in plain sentences |
640
1077
 
@@ -654,9 +1091,11 @@ Options: `issuerKeys` (the thumbprints of issuer keys you trust),
654
1091
  expect), `stampServices` (the fingerprints of the certificates of the
655
1092
  time-stamp services you trust), `blocks` (the fingerprints of the
656
1093
  blocks of a public blockchain you trust), `vouchers` (the fingerprints of the key
657
- sets of the organisations whose vouching you trust), `disclosures` (the
1094
+ sets of the organisations whose vouching you trust), `witnesses` (the
1095
+ fingerprints of the key sets of the witnesses you trust), `disclosures` (the
658
1096
  disclosures of covered fields that you were handed), `cancellations`
659
- (the person's own copies of cancellations), `expectedRoot` and
1097
+ (the person's own copies of cancellations), `seals` (seals of the book
1098
+ held elsewhere), `expectedRoot` and
660
1099
  `expectedSize` (a top fingerprint and a number of entries you already
661
1100
  hold a copy of). Give
662
1101
  them whenever you have them.
@@ -680,7 +1119,8 @@ Writing records: `prepareSlip` and `assembleSlip`, `prepareApproval` and
680
1119
  `writeRefusal` and `writeTerms` (the service), `writeVouching` and
681
1120
  `writeWithdrawal` (an organisation), `writePass` (an agent, to a helper
682
1121
  agent), `writeSeal` (whoever keeps
683
- the book; it needs Node.js), `writeBook` and `makeShow`, all from
1122
+ the book; it needs Node.js), `writeWitness` and `openWitness` (a
1123
+ witness; `openWitness` needs Node.js), `consistencyProof`, `writeBook` and `makeShow`, all from
684
1124
  `provared`. The library asks no time-stamp service itself: you send
685
1125
  the seal's fingerprint to the service you choose and put its answer
686
1126
  beside the seal. In a browser, `createPasskey`, `signSlipWithPasskey`,
@@ -706,6 +1146,10 @@ prepared, then `assembleSlip`; keep the disclosures it prepared.
706
1146
  - A name in a record is a label. Only keys are checked. A vouching
707
1147
  record shows which organisation stood behind a name, not how well it
708
1148
  checked.
1149
+ - An identifier of another system's, named in a stub, shows that the
1150
+ two records speak of one call, not that the call did what either
1151
+ says. A fingerprint of a request, an answer, the arguments or an error
1152
+ shows only that the content matches, not that it was the same call.
709
1153
 
710
1154
  And about time:
711
1155
 
@@ -717,7 +1161,10 @@ And about time:
717
1161
  the agent outside its slip;
718
1162
  - the last seal, with everything after it, can itself be removed. Only a
719
1163
  copy of it, or of its top fingerprint and number of entries, held by
720
- someone else shows that;
1164
+ someone else shows that: hand such a copy of a seal to the checker
1165
+ (`--seal`, or `seals`), such as the latest seal a witness holds;
1166
+ - a witness statement is the witness's own word for when it saw the
1167
+ seal, and is not used as a time-stamp;
721
1168
  - time-stamp services sign with methods of today's kind, which are not
722
1169
  quantum-safe.
723
1170
 
@@ -730,7 +1177,9 @@ connects to anything: to the Bitcoin nodes you name, for block headers.
730
1177
 
731
1178
  ## Not built yet
732
1179
 
733
- A public place where seals are kept; covering fields other than a slip's
1180
+ A public place where seals are kept, which gathers the seals of many
1181
+ books; a witness that runs as a service (the library holds the part that
1182
+ decides and signs); covering fields other than a slip's
734
1183
  names and purpose; connectors for particular agent software (the
735
1184
  general one, `recordTools`, is built); a compact binary form; the
736
1185
  recorder service. The checking page cannot yet take a chain of block