[Alex-Dolid]

Summary

A new docs page compares syncom with RxDB 17.6.0 in 300 rows. Deciding the
syncom column from running code turned up 12 bugs in syncom. Each is fixed
here with a regression test, and the cells describe the code after the fixes.

Docs: /guide/vs-rxdb

• Scope. 300 rows in eight sections: package and platform, storage, data
  model and writes, queries and reactivity, replication, security and
  integrity, extensibility and operations, project and ecosystem. A summary,
  the method, an RxDB → syncom concept map, speed charts and "which to
  choose" go with it.
• How the RxDB column was decided. From the source at the 17.6.0 tag,
  from runs of the published package (with and without dev-mode), and from
  all 271 pages of rxdb.info. Two adversarial passes tried to refute the
  strongest claims about each project; none was refuted, and 13 were reworded.
• Premium. Premium features are marked ⚠️ with their tier. They are
  closed source, so those cells follow the docs.
• Rendering. docs/comparison-rxdb.json is rendered by
  scripts/comparison.js, which now renders both matrices. A group's
  section becomes its own ### table. .sc-matrix-duel styles the
  two-column table.
• Charts. BenchCompare / BenchFigure accept only (library ids), and
  the model functions take an optional filter.
• Main matrix. docs/comparison.json gets the RxDB and syncom cells the
  deeper reading corrected, and comparison.md / why.md no longer claim a
  syncom host stops a write its validator rejects (it relays it).

Fixes

Security:

• Scopes bound what a peer writes, not only what it declares.
  • An update() took the scope of its delta. With a field-based
    sync.scope it was minted global and reached every peer. An op now
    takes the scope of its record as it leaves it, re-read after validation.
  • A host trusted the scope an op declared, so a peer granted Other
    overwrote a Doc record. A host's push now passes the grant to
    applyRemote as bound.
  • An op whose record is outside the grant, before or after the op, is now
    refused: outOfScope in the result, a refused event with
    reason: 'scope'.

Replication:

• Offline writes past retention.ops reach the host. A dialing node
  keeps its own ops until every enabled registered node holds them
  (OpLog.retain(keep); acked in the registry, shown by listNodes()).
  A host that lacks ops the log no longer holds is reported once per hello
  as UNDELIVERABLE.
• A restored directory no longer forks its node in silence. At hello the
  node notices the host holds more of its own stream, and reports REWOUND.
  It then moves past the host's seqs and bootstraps back what the lost ops
  wrote.

Storage:

• Logged is committed. Before, a record file (or chain block) that failed
  after the log append rejected the write, though its op stayed logged and
  replicated. Now the write stands, the entry is held in memory, the batch is
  replayed onto the files at the next open, and UNPERSISTED is reported. A
  peer's batch is held the same way.
• A torn record file is INTEGRITY, not a bare SyntaxError.

API:

• Collections: a collection's get / update / delete / record act
  only on its own records (null / NOT_FOUND otherwise). An entity named
  like a storage member (get, close, log…) is refused at open().
• Queries: cursor.one() no longer leaves limit(1) on the cursor.
  Indexed and unique object values match whatever their key order
  (canonical keyOf).
• CRDT helpers: merge() / diff() walk own keys only and never take
  __proto__.
• Node.js ESM: import { open } from '@alexify/syncom' works. Both
  barrels spell their exports out so cjs-module-lexer can read them.

Docs drift:

• Failing listeners and watch runs log at warn, not debug.
• An op that skips ahead is dropped as a gap, not set aside.
• There is no read-only open for a second tab.
• A host relays an op its validator or beforeApply skipped.
• A nested object is one register.
• The event tables list log, refused and every error code.
• sync.op.outOfScope, which never existed, is gone from the table.

Behaviour changes

• A record-file failure after the log append no longer rejects the write: it
  emits error with UNPERSISTED.
• collection.get(id) of another collection's record returns null;
  update / delete / record reject with NOT_FOUND.
• The applyRemote result and the apply / sync event payloads gain
  outOfScope.
• New error codes: UNPERSISTED, REWOUND, UNDELIVERABLE. They are added
  to the SyncomErrorCode union and the codes table in api.md.
• Browser budget 41 → 42 KB. The fixes add about 1.2 KB min+gzip; the
  entry is 41.9 KB. Size figures in the README and docs are re-measured
  (pnpm size, pnpm bench:size, pnpm bench:report).

Not covered

• Ops written before any node was registered, and compacted since, still
  cannot reach a host. This is now reported (UNDELIVERABLE), not silent.
• pnpm bench:compare was not re-run; only the size figures are re-measured.
• RxDB Premium cells rest on its docs (closed source).

Testing

• pnpm test: 752 tests, 0 failures.
• pnpm test:coverage: 99.6 / 96.4 / 99.9 / 99.6 (stmts / branches / funcs /
  lines).
• pnpm test:types, pnpm lint, pnpm format:check, pnpm size and
  pnpm docs:build (no dead links) all pass.
• New tests: tests/sync.retention.test.js (own-op retention,
  UNDELIVERABLE, REWOUND), tests/package/esm.test.js (ESM named
  imports of every entry), plus additions to sync.scopes, storage.fsi,
  storage, crdt and the query suite.
• I also re-ran the scripts that reproduced the scope bugs; both now show the
  bug gone.
• I opened /guide/vs-rxdb in the browser at desktop and phone widths.

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
syncom	Ready	Preview	Oct 23, 2026 8:08pm UTC