Initial Commit

openspec/archive/2026-05-04-network-plugin-state/proposal.md

@@ -0,0 +1,96 @@
+# OpenSpec Change: network-plugin-state
+
+- **Change ID**: network-plugin-state
+- **Package**: `@suaveplan/network-plugin-state` (new)
+- **Status**: PROPOSED
+- **Type**: new-package
+- **Date**: 2026-04-29
+
+## Why
+
+Multiplayer games and collaborative applications require synchronized state across all peers in a mesh. Without a dedicated state synchronization layer, each application must implement its own conflict resolution, delta compression, and reconnection handling. This leads to inconsistent state, bandwidth waste, and race conditions. The genesis `bitpacker` package provides efficient binary delta compression, and `serialize` provides schemas. This plugin implements a mesh-wide state store with automatic conflict resolution.
+
+This plugin implements **synchronized state with delta compression, path-specific conflict resolution, and Bloom filter comparison** for efficient sync.
+
+## What changes
+
+- **ADDED**: `StatePlugin` class implementing `NetworkPlugin`.
+- **ADDED**: `StateManager` — maintains a versioned local state store with path-based access (`get(path)`, `set(path, value)`). Tracks state history with sequence numbers. Supports subscriptions per path.
+- **ADDED**: `DeltaCompressor` — computes deltas between state versions using `@suaveplan/bitpacker`. Only changed properties are transmitted. Supports schema-aware compression for typed data.
+- **ADDED**: `StateReconciler` — resolves conflicts between concurrent state updates with 5 strategies:
+  - `latest-wins` — highest timestamp wins
+  - `remote-wins` — always accept remote state
+  - `local-wins` — always keep local state
+  - `deep-merge` — combine non-conflicting properties recursively
+  - `custom` — application-provided merge function
+- **ADDED**: `PathStrategyRegistry` — applies different reconciliation strategies per state path. Example: `game/positions` uses `latest-wins`, `game/inventory` uses `deep-merge`, `game/achievements` uses `remote-wins`.
+- **ADDED**: `BloomFilterComparator` — uses Bloom filters for efficient state comparison. Peers exchange compact Bloom filter summaries to detect divergence without transmitting full state.
+- **ADDED**: `StateSyncEngine` — orchestrates sync: detect divergence (Bloom filter), request delta, apply delta, resolve conflicts, notify subscribers. Supports full-state sync for new peers and delta sync for existing peers.
+- **ADDED**: `ResourceAdapter` — adapts sync behavior based on device profile:
+  - `EDGE` — minimal history, aggressive compression, low sync frequency
+  - `MOBILE` — balanced compression, medium sync frequency, battery-aware
+  - `DESKTOP` — full history, fast sync, high responsiveness
+  - `CLOUD` — maximum throughput, aggressive batching
+- **ADDED**: `StateHistoryPruner` — prunes old history based on memory constraints and policy. Configurable retention window.
+- **ADDED**: `NETWORK_STATE_CODES` — frozen error registry.
+
+## State sync flow
+
+```text
+Peer A changes state at path "game/positions/player1"
+  ↓
+StateManager records change with sequence + timestamp
+  ↓
+DeltaCompressor computes delta from previous version
+  ↓
+StateSyncEngine broadcasts delta to mesh peers
+  ↓
+Peer B receives delta
+  ↓
+StateReconciler applies strategy for "game/positions/*" (latest-wins)
+  ↓
+StateManager updates local state, notifies subscribers
+```
+
+## Out of scope
+
+- Game-specific entity replication (ECS) → `gaming-ecs`, `gaming-r3f-networking`.
+- Persistent state storage → future `network-plugin-state-persistence`.
+- CRDT implementations → future `network-plugin-state-crdt`.
+
+## Source mapping
+
+This derives from `docs/old-network-planning/PHASE_7.md` §State Management and `PLAN_ENHANCED_PHASE_7.md`.
+
+| Requirement | Source lines | Notes |
+|---|---|---|
+| State history tracking | `PHASE_7.md` §StateManager | reworded; path-based instead of object-based |
+| Delta compression | `PHASE_7.md` §BitPackerCompressor | reworded; uses `@suaveplan/bitpacker` |
+| Reconciliation strategies | `PHASE_7.md` §StateReconciler | verbatim: 5 strategies |
+| Path-specific resolution | `PHASE_7.md` §Path-specific Resolution | verbatim |
+| Bloom filter comparison | `ADAPT_3.md` §Performance Optimizations | reworded; for state divergence detection |
+| Resource profiles | `PLAN_ENHANCED_PHASE_7.md` §Resource Profiles | reworded; 4 profiles |
+
+### Requirements with no source line
+
+| Requirement | Justification |
+|---|---|
+| `StateSyncEngine` orchestration | Required to automate the full sync pipeline |
+| `StateHistoryPruner` | Required to prevent unbounded memory growth |
+| Subscriptions per path | Required for reactive UI updates |
+
+## Verification approach
+
+- `src/state/state.spec-conformance.test.ts` — pins get/set, path access, history, subscriptions.
+- `src/delta/delta.spec-conformance.test.ts` — pins delta computation, roundtrip, schema encoding.
+- `src/reconciler/reconciler.spec-conformance.test.ts` — pins all 5 strategies, path-specific application.
+- `src/sync/sync.spec-conformance.test.ts` — pins divergence detection, delta request, full-state sync.
+- `src/bloom/bloom.spec-conformance.test.ts` — pins filter comparison, false-positive tolerance.
+- `src/resource/resource.spec-conformance.test.ts` — pins profile adaptation, compression level, sync frequency.
+- `src/pruner/pruner.spec-conformance.test.ts` — pins retention window, memory bounds.
+- `src/error-codes.spec-conformance.test.ts` — frozen registry.
+- `tests/dep-graph.test.ts` — pins direct import of `@suaveplan/bitpacker`.
+
+## Migration impact
+
+None — additive only.