How do you design offline support and synchronization?

Offline support is a data design problem before it is a UI problem. Reads need a cache with clear freshness rules. Writes need a durable outbox so they survive process death. The sync engine needs ordering, retries and conflict handling, and the UI needs honest states for pending, synced and failed.

class Outbox {
  final List<PendingAction> _pending = [];

  Future<void> enqueue(PendingAction action) async {
    _pending.add(action);
    await _flush();
  }

  Future<void> _flush() async {
    while (_pending.isNotEmpty) {
      final action = _pending.first;
      try {
        await api.send(action);
        _pending.removeAt(0);
      } on NetworkException {
        return;
      }
    }
  }
}

• Persist the outbox, not only an in-memory queue; the app may be killed
• Give each mutation a client-generated id so retries are idempotent
• Define freshness per resource rather than one global cache duration
• Show pending and failed writes in the UI, with a retry action
• Handle conflicts with last-write-wins, version checks or field-level merge, and document the choice

Detecting connectivity is not enough

The device can report a connection while the backend is unreachable, rate limited or behind a captive portal. Treat connectivity as a hint and let actual request outcomes drive sync. Back off when failures repeat instead of hammering the server.