@rljson/network 0.0.1

package/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2025 Rljson
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package/README.architecture.md

@@ -0,0 +1,290 @@
+<!--
+@license
+Copyright (c) 2025 Rljson
+
+Use of this source code is governed by terms that can be
+found in the LICENSE file in the root of this package.
+-->
+
+# Architecture
+
+## Design Principle: Zero RLJSON Knowledge
+
+The network package knows **nothing** about Io, Bs, Db, trees, hashes, or
+sync. It only knows about nodes, peers, and topology. This makes it reusable
+for any application that needs self-organizing star topology.
+
+- Use `domain` (network grouping), **never** `treeKey` (RLJSON concept)
+- No imports from `@rljson/io`, `@rljson/bs`, `@rljson/db`, `@rljson/server`
+- The bridge between network → RLJSON lives in `@rljson/server`, not here
+
+## Module Structure
+
+```
+src/
+├── types/
+│   ├── node-info.ts            // NodeId, NodeInfo — core identity type
+│   ├── peer-probe.ts           // PeerProbe — probe result
+│   ├── network-topology.ts     // NetworkTopology, NodeRole, FormedBy
+│   ├── network-config.ts       // Config interfaces for all discovery layers
+│   └── network-events.ts       // Event types emitted by NetworkManager
+├── identity/
+│   └── node-identity.ts        // Persistent UUID, hostname, IP discovery
+├── election/
+│   └── hub-election.ts         // Deterministic hub election algorithm
+├── probing/
+│   ├── peer-prober.ts          // Real TCP connect probe via node:net
+│   └── probe-scheduler.ts      // Periodic probing + change detection
+├── layers/
+│   ├── discovery-layer.ts      // DiscoveryLayer interface — contract for all layers
+│   ├── broadcast-layer.ts      // BroadcastLayer — UDP broadcast discovery (Try 1)
+│   ├── manual-layer.ts         // ManualLayer — always-present override
+│   └── static-layer.ts         // StaticLayer — hardcoded hub fallback (Try 3)
+├── peer-table.ts               // PeerTable — merged view of peers from all layers
+├── network-manager.ts          // NetworkManager — central orchestrator
+├── example.ts                  // Runnable example
+└── index.ts                    // Public API exports
+```
+
+## Fallback Cascade
+
+Discovery layers are tried in order of autonomy:
+
+| Position     | Layer         | Required?          | Description                              |
+| ------------ | ------------- | ------------------ | ---------------------------------------- |
+| **Try 1**    | UDP Broadcast | **always on**      | Zero-config LAN discovery. Primary path. |
+| **Try 2**    | Cloud Service | **optional**       | Cross-network fallback.                  |
+| **Try 3**    | Static Config | **optional**       | Hardcoded `hubAddress`. Last resort.     |
+| **Override** | Manual / UI   | **always present** | Human escape hatch. No config entry.     |
+
+## Hub Election Rules
+
+1. **Incumbent advantage**: If a hub exists and is reachable, keep it.
+2. **Earliest `startedAt`**: If no incumbent, the node started first wins.
+3. **Tiebreaker**: Lexicographic `nodeId` comparison.
+
+### Election Implementation
+
+`electHub()` in `src/election/hub-election.ts` is a **pure function** — no
+I/O, no side effects, fully deterministic:
+
+- **Input**: candidates (`NodeInfo[]`), probes (`PeerProbe[]`), current hub, self
+- **Output**: `ElectionResult` with `hubId` and `reason`
+- Self is **always** considered reachable (no need for self-probe)
+- Returns `{ hubId: null, reason: 'no-candidates' }` when no reachable nodes
+- Reasons: `'incumbent'`, `'earliest-start'`, `'tiebreaker'`, `'no-candidates'`
+
+## Probing Design
+
+### PeerProber
+
+`probePeer()` in `src/probing/peer-prober.ts` uses real TCP connect via
+`node:net`:
+
+- `connect({ host, port, timeout })` — non-blocking socket connect
+- Never rejects — always resolves with a `PeerProbe` result
+- `reachable: true` on `'connect'`, `false` on `'timeout'` or `'error'`
+- Latency measured with `performance.now()` (sub-millisecond precision)
+- Socket is destroyed immediately after result (no keep-alive)
+
+### ProbeScheduler
+
+`ProbeScheduler` in `src/probing/probe-scheduler.ts` manages periodic
+probing of all known peers:
+
+- **Injectable probe function** (`ProbeFn`) for Tier-1 mock tests
+- **Real TCP** by default (uses `probePeer`)
+- **Self-exclusion at probe time** — filters self in `_runCycle()`, not
+  `setPeers()`, because `setPeers()` may be called before `start()` sets
+  the self ID
+- **Change detection**: tracks `_wasReachable` state per peer, emits
+  `'peer-unreachable'` and `'peer-reachable'` only on state transitions
+- **Flap dampening**: a peer must fail `failThreshold` consecutive probes
+  (default: 3) before being declared unreachable. A single success resets
+  the counter immediately. This prevents flapping on transient network glitches.
+- **No false alerts**: first probe sets baseline, subsequent probes detect
+  changes
+- `runOnce()` — manual single cycle for deterministic test control
+- Events: `'probes-updated'`, `'peer-unreachable'`, `'peer-reachable'`
+
+## Testing Strategy (3-Tier)
+
+| Tier   | Scope                  | What's real        | Used for                             |
+| ------ | ---------------------- | ------------------ | ------------------------------------ |
+| Tier 1 | Unit (mock probes)     | Logic + state only | Election, scheduling, event emission |
+| Tier 2 | Real TCP (localhost)   | Actual socket I/O  | Probing, latency, connection refused |
+| Tier 3 | Multi-process (future) | Full network stack | True distributed scenarios           |
+
+- Tier 1 uses `ProbeFn` injection for deterministic, fast tests
+- Tier 2 uses `node:net.createServer()` on localhost with random ports
+- All tiers run in the same `pnpm test` suite
+
+## NodeIdentity Design
+
+`NodeIdentity` uses dependency injection for full testability:
+
+- `NodeIdentityDeps` interface — all OS/fs functions injectable
+- `defaultNodeIdentityDeps()` — real Node.js implementations
+- `NodeIdentity.create(options)` — async factory, loads or generates UUID
+- UUID persisted at `<identityDir>/<domain>/node-id`
+- Same machine + same domain = same identity across restarts
+
+## DiscoveryLayer Contract
+
+All discovery mechanisms implement the `DiscoveryLayer` interface:
+
+```
+start(identity) → Promise<boolean>   // Returns false if layer can't operate
+stop()          → Promise<void>      // Clean up resources
+isActive()      → boolean            // Whether currently running
+getPeers()      → NodeInfo[]         // Currently known peers
+getAssignedHub() → string | null     // Hub dictated by this layer
+on/off          → event subscription // peer-discovered, peer-lost, hub-assigned
+```
+
+### ManualLayer
+
+- **Always present**, cannot be disabled
+- Does **not** discover peers — only overrides hub assignment
+- `assignHub(nodeId)` / `clearOverride()` API
+- Clearing returns control to the automatic cascade
+
+### BroadcastLayer (Try 1 — Primary Discovery)
+
+UDP broadcast discovery for zero-config LAN peer detection:
+
+- **Self-test on startup**: sends a broadcast and waits for loopback reception;
+  if the self-test times out, broadcast is unavailable and the cascade falls
+  through to the next layer (cloud or static)
+- **Periodic broadcasting** at `intervalMs` (default: 5000 ms)
+- **Domain filtering**: ignores packets from nodes in a different `domain`
+- **Self-packet detection**: loopback packets are silently dropped (never
+  added to the peer table)
+- **Peer timeout**: peers that haven't been heard from within `timeoutMs`
+  (default: 15000 ms) are removed and `peer-lost` is emitted
+- **Does NOT assign hub** — `getAssignedHub()` always returns `null`.
+  Hub is elected by the NetworkManager via `electHub()` when broadcast
+  peers are available (`formedBy: 'broadcast'`)
+- **Testable via `UdpSocket` interface** — injectable socket factory
+  (`CreateUdpSocket`) for deterministic mock-based testing with
+  `MockUdpHub` / `MockUdpSocket`
+
+Configuration (`BroadcastConfig`):
+
+```typescript
+{
+  enabled: boolean;      // default: true
+  port: number;          // default: 41234
+  intervalMs?: number;   // default: 5000
+  timeoutMs?: number;    // default: 15000
+}
+```
+
+### StaticLayer
+
+- Last resort fallback (Try 3)
+- Reads `hubAddress` from config (`"ip:port"`)
+- Creates a **synthetic peer** with deterministic nodeId `static-hub-<address>`
+- Returns `false` from `start()` if no `hubAddress` configured
+- Emits `peer-discovered` + `hub-assigned` on start, `peer-lost` on stop
+
+## PeerTable Design
+
+Merged view of all peers from all discovery layers:
+
+- **Deduplication by nodeId** — same peer from multiple layers appears once
+- **Per-layer tracking** via `_layerPeers` map for correct removal semantics
+- `peer-joined` fires only when a genuinely **new** peer is first seen
+- `peer-left` fires only when **all** layers have lost the peer
+- `setSelfId()` excludes own node from the peer table
+- `attachLayer()` imports existing peers + subscribes to future events
+
+## NetworkManager Design
+
+Central orchestrator — the main public API:
+
+- Creates `NodeIdentity` on start (respects `identityDir` from config)
+- Starts ManualLayer, BroadcastLayer, CloudLayer, and StaticLayer in cascade order
+- Creates and manages `ProbeScheduler` for reachability checking
+- Uses `PeerTable` for merged peer tracking
+- Applies **cascade logic** via `_computeHub()`:
+  1. Manual override wins
+  2. Election via probes (if probes available → `electHub()`)
+     - `formedBy: 'broadcast'` when broadcast layer is active with peers
+     - `formedBy: 'election'` otherwise
+  3. Cloud assignment (cloud dictates hub — has the full picture)
+  4. Static config fallback
+  5. No result → `unassigned`
+- Accepts `NetworkManagerOptions` with injectable `probeFn`,
+  `broadcastDeps`, and `cloudDeps` for testing
+- Emits events: `topology-changed`, `role-changed`, `hub-changed`,
+  `peer-joined`, `peer-left`
+- Continuous re-evaluation: any peer/hub change or probe update triggers
+  `_recomputeTopology()`
+- `getProbeScheduler()` provides public access for advanced usage
+
+## Known Limitations (Future Work)
+
+The current election system is designed for **LAN office sync** (2–10 nodes)
+and works well in that scope. The following distributed-systems gaps are
+documented for future epics:
+
+### Split-Brain — No Consensus Protocol
+
+Each node runs its own `electHub()` independently. In a network partition,
+two isolated groups may each elect a different hub. There is no Raft, Paxos,
+or gossip protocol to reach agreement.
+
+**Mitigation**: On LAN, partitions are rare. When the partition heals, probes
+converge and nodes re-elect the same hub (incumbent advantage + earliest
+`startedAt`).
+
+**Future**: A gossip protocol (Epic 5+) or Raft-based consensus could
+guarantee a single hub across partitions.
+
+### No Quorum
+
+Election doesn't require a majority of nodes to agree. A single node
+in isolation will elect itself as hub. This is by design for small LANs
+but may be undesirable in larger deployments.
+
+**Future**: Add optional `quorum: true` config that requires `>50%` of
+known peers to be reachable before accepting an election result.
+
+### Probe Staleness Window
+
+The default probe interval is 10 seconds. During this window, a hub can
+go down without being detected. This is acceptable for file sync but not
+for real-time systems.
+
+**Mitigation**: `intervalMs` is configurable. For faster detection, reduce
+the interval (at the cost of higher network traffic).
+
+**Future**: Hub heartbeat / lease mechanism — the hub actively sends
+keepalives; clients consider the hub down if heartbeat stops.
+
+### No Handover Protocol
+
+When a hub is replaced, there is no graceful handover of state (e.g.,
+pending syncs, in-flight messages). Clients simply reconnect to the new hub.
+
+**Future**: Handover protocol lives in `@rljson/server`, not in the
+network layer. The network layer only signals topology changes; the
+application layer (`@rljson/server`) must handle state transfer.
+
+### No Hub Lease / Heartbeat
+
+There is no mechanism for the hub to "renew" its leadership. If the hub
+stops responding but probe timing creates a race, two nodes might briefly
+disagree about who is hub.
+
+**Mitigation**: Flap dampening (3 consecutive failures required) prevents
+transient disagreements. Incumbent advantage prevents unnecessary re-elections.
+
+**Future**: A time-limited hub lease (e.g., 30 seconds) with active renewal
+would provide stronger guarantees.
+
+## Layer 0 — No Dependencies
+
+This package has **zero** `@rljson/*` dependencies. It sits at Layer 0
+alongside `@rljson/rljson`, meaning it can be published independently.

package/README.blog.md

@@ -0,0 +1,11 @@
+<!--
+@license
+Copyright (c) 2025 Rljson
+
+Use of this source code is governed by terms that can be
+found in the LICENSE file in the root of this package.
+-->
+
+# Blog
+
+Add latest posts at the end.

package/README.contributors.md

@@ -0,0 +1,32 @@
+<!--
+@license
+Copyright (c) 2025 Rljson
+
+Use of this source code is governed by terms that can be
+found in the LICENSE file in the root of this package.
+-->
+
+# Contributors Guide
+
+- [Prepare](#prepare)
+- [Develop](#develop)
+- [Administrate](#administrate)
+- [Fast Coding](#fast-coding)
+
+## Prepare
+
+Read [prepare.md](doc/prepare.md)
+
+<!-- ........................................................................-->
+
+## Develop
+
+Read [develop.md](doc/develop.md)
+
+## Administrate
+
+Read [create-new-repo.md](doc/create-new-repo.md)
+
+## Fast Coding
+
+Read [fast-coding-guide.md](doc/fast-coding-guide.md)

package/README.md

@@ -0,0 +1,24 @@
+<!--
+@license
+Copyright (c) 2025 Rljson
+
+Use of this source code is governed by terms that can be
+found in the LICENSE file in the root of this package.
+-->
+
+# @rljson/network
+
+## Users
+
+| File                                 | Purpose                     |
+| ------------------------------------ | --------------------------- |
+| [README.public.md](README.public.md) | Install and use the package |
+
+## Contributors
+
+| File                                             | Purpose                       |
+| ------------------------------------------------ | ----------------------------- |
+| [README.contributors.md](README.contributors.md) | Run, debug, build and publish |
+| [README.architecture.md](README.architecture.md) | Software architecture guide   |
+| [README.trouble.md](README.trouble.md)           | Errors & solutions            |
+| [README.blog.md](README.blog.md)                 | Blog                          |