Two batches, verified against nextgraph-rs throughout.
P1a — the capability surface. Reading was an ACL (Map<doc, Set<principal>>),
the exact inversion of key possession. It is now possession: `capFor(nuri)` is
the only question, there is no principal parameter anywhere, and nothing turns
a bare reference into a cap. Sharing is `shareCap(cap, toInbox)`, a Link
deposit; receiving needs no operation. `Nuri` and `ReadCap` are template
literal types, so passing a bare reference where a cap belongs is a compile
error, with runtime guards behind it for JavaScript callers.
The virtual user boundary. Every access function is now confined to the
connected user, through two rules on one criterion (possession), implemented in
two places so a lapse in either is caught by the other: authorization at the
passage points, and "do not even attempt" at the callers. The polyfill's own
machinery moved to physical.ts — unguarded, never exported — which replaced an
exemption list: the machinery no longer gets waved through the guard, it calls
something the guard never saw.
Removed, as emulating capabilities the target does not have:
- discovery.ts and its global index. There is no discovery in NextGraph; you
follow links. It also pooled user data across wallets.
- the cross-account fan-out (listEntityDocs, resolveReadGraphs, allAccounts,
loadShim), which was cross-user enumeration by construction.
- resolveInboxAnchor, a single inbox common to every user.
Caps are now stored where NextGraph stores them, and read back rather than
recomputed: AddRepo on the store's Store branch for documents a user creates,
AddLink on its User branch for caps received. Inboxes belong to someone — the
user's own, plus one per document — and connecting a user drains them all;
that is the library's job, not the app's.
Corrections worth recording: a ReadCap is `r:`, not `:k:` (reported by
NextGraph's developer, verified in BlockRef::readcap_nuri); received caps DO
have a register (AddLink), contrary to what this repo's notes claimed; and
"wallet" upstream means keyring — what owns three stores is a user, so the
vocabulary follows.
The cap value is the constant OK: the only question the emulation answers is
whether a cap is held. P1b replaces that one constant with a real key.
After this the shape is right and the isolation is still fake. Nothing here may
be described as anonymous or private.
Bug: à la reconnexion, resolveAccount lisait le shim depuis le store-root
(did🆖${privateStoreId}), NON abonnable → pas de barrière first-State → un "0 rows"
à froid est ambigu → le retry (resolveAccountReliably/provisionRetry) échoue → nouveau
compte provisionné → FORK → données du compte invisibles.
Cause NextGraph (vérifiée nextgraph-rs): "trouvable-sans-lookup" (store-root) et
"abonnable" (did:ng:o:<RepoID aléatoire>) sont DISJOINTS — pas de doc à la fois
devinable et attendable → une résolution shim purement barrière est impossible.
Fix (indirection pointeur → doc-shim abonnable):
- Les AccountRecord migrent dans un doc-shim doc_create'd (did:ng:o:..., a une barrière).
- Un pointeur écrit-une-fois dans le store-root (<shim:root> <shim:shimDoc> <docShim>)
le nomme. resolveShimDoc lit le pointeur → ensureRepoOpen(docShim) [barrière] → lecture
de compte AUTORITATIVE (cold 0 = absent pour de vrai). Retry de compte SUPPRIMÉ.
- Micro-garde résiduel (pointerGuard, ex-provisionRetry) sur le SEUL triple pointeur
écrit-une-fois; ne peut jamais forker un compte; fork de pointeur réconcilié au
doc-shim canonique (lexicographiquement-min), sans perte.
- Migration: migrateLegacyRecords copie (pas déplace) les comptes de l'ancien store-root
vers le doc-shim avant toute conclusion "absent"; idempotent; wallet neuf → no-op.
Tests: unit 128/128, e2e réel 42/42 (CONTRACT 2 = non-fork du compte à la reconnexion),
red-before/green-after prouvé. Docs: nextgraph-current-state (antagonisme + indirection),
simulation, migration-guide.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Les fix récents (anti-fork, open-repo, access-log) n'avaient AUCUN test de
comportement dans le domaine de la lib. Ajout de tests DÉTERMINISTES à base de `ng`
factice modélisant la condition d'échec (pas de broker réel — le harness e2e réel
réhydrate trop vite et ne reproduit pas ces cas).
- anti-fork.test.ts : fake avec lag de sync (0 lignes les K premières lectures du
record de compte, puis les vraies). Asserte le cœur du fix : compte retrouvé au
retry → réutilisé, 0 doc_create (pas de fork) ; compte réellement neuf → 1 seul
jeu de docs provisionné, budget de retry prouvé consommé ; idempotence de session ;
cas limite « trouvé à la dernière tentative ».
- open-repo.test.ts : fake où une requête ancrée rend vide tant que doc_subscribe
n'a pas ouvert le repo. Asserte subscribe-avant-read + idempotence (Set des repos
ouverts) + no-op si le ng n'a pas doc_subscribe.
- access-log.test.ts : off par défaut (silencieux), on via config ET via env
NG_EVENTUALLY_ACCESS_LOG, préfixe = identité active (suit setCurrentUser),
row-count sur READ. Restaure console.log/env fidèlement.
bun test : 91 → 117 pass, 0 fail. tsc --noEmit propre. src/ non touché.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Sylvain Duchesne