Real-time Data Sync

Implement real-time sync using operation-based updates, versioning, and conflict-free patterns. This guide focuses on architecture, not a single CRDT implementation.

1) Pick a sync model

• Operation-based: exchange operations (best for real-time UX)
• State-based: exchange full snapshots periodically (simpler, more bandwidth)
• Hybrid: ops for live edits + snapshots for recovery

2) Use a dedicated channel

p2p.onMessage { msg ->
    if (msg.channel != "sync") return@onMessage
    // decode and apply operation
}

3) Version and de-duplicate

Include an operation ID and a monotonic counter per author. Track seen ops to avoid loops.

{
  "v": 1,
  "type": "sync.op",
  "doc": "doc-123",
  "opId": "peer-9:1042",
  "author": "peer-9",
  "payload": { /* op */ }
}

4) Recovery and catch-up

• On reconnect, request missing ops since a known watermark.
• If the gap is too large, fallback to snapshot sync.

5) Security + performance

• Authenticate peers (see Security Best Practices).
• Bound op rates and payload sizes.
• Batch ops to reduce overhead (see Performance Optimization).