Analyze @tomic/lib vs atomic_lib duplication (and WASM tradeoffs)

[joepio]

Related Issues

Analysis request: how much logic is duplicated between @tomic/lib and atomic_lib / WASM, and whether migrating the TS library onto WASM would help.

Summary

There are two different overlaps, and they should not be treated as one problem:

1. Must-match protocol (~1.5–2.5k lines of TS): genesis certificates, commit canonical JSON, WS v2 frames, RBSR fingerprints, drive hashes, datatype tags. These must stay byte-identical with Rust. They are already pinned with golden vectors and run off the render path.
2. Two embeddings of a Resource: TS Resource/Store is a React cache + outbox + TipTap LoroDoc. Rust Resource/Db is a durable node. Flutter already skipped the TS embedding and calls atomic_lib directly.

Recommendation: do not move Resource.get / the JS cache into WASM. Continue the existing AtomicNode / ClientDb direction for query, persist, and apply-commit.

Full write-up: planning/ts-wasm-duplication.md.

Measured costs (in-process; Worker would be slower)

Op	Time
JS Resource.get(name)	~0.07 µs
WASM getResource → JSON-AD	~75 µs (~1000×)
WASM putResource (parse + index)	~575 µs
WASM blake3 1 KiB	~1.4 µs (faster than noble JS ~16 µs)
Native Ed25519 sign	~24 µs vs noble JS ~284 µs

WASM binaries already loaded by the browser: atomic_wasm 6.1 MB (2.3 MB gzip) + loro-crdt 3.1 MB (1.0 MB gzip). TipTap needs a main-thread loro-crdt LoroDoc; folding Resource into atomic_wasm does not remove that second heap.

How to reproduce

cargo run -p atomic_lib --example protocol_microbench --release
cd browser/lib && pnpm bench src/ts-vs-wasm.bench.ts   # needs wasm/pkg

Checklist

• Add changelog entry linking to issue, describe API changes — N/A (planning + benches only)
• Add or update tests if needed — added reproducible benches, not product tests
• Update docs if needed — planning/ts-wasm-duplication.md + README index