CDUS Architecture & Core Engine

CDUS is intentionally engineered without heavy asynchronous runtimes like Tokio to maintain absolute predictability, zero thread-pool starvation, and minimal memory footprint on resource-constrained mobile and embedded environments.

Design Philosophy

• Thread-Pipelined Concurrency: Long-running system stages run as dedicated OS threads communicating across bounded Flume channels.
• Protocol Stability: All wire communications and internal IPC messages serialize using MessagePack for fast packing and unpacking.
• Zero Heavy Async: Disk I/O, socket polling, and cryptographic operations run on dedicated pipelines, preventing transfer bottlenecks from stalling real-time clipboard notifications.

System Pipeline Diagram

[Network Inbound Socket]
          │
          ▼
   (Decrypt & Verify) ──> Noise XX / IK Session Decryptor
          │
          ▼
   [Inbound Message Dispatcher]
          ├───────────────┬───────────────┐
          ▼               ▼               ▼
   (Clipboard Queue) (File Chunk Pipeline) (Heartbeat / PEX)
          │               │               │
          ▼               ▼               ▼
     OS Clipboard     Disk Writer       Routing Table

Transport Architecture

CDUS employs a dual-path networking engine:

1. Direct Peer-to-Peer (LAN & PEX): Discovers local peers using mDNS UDP broadcasts and Peer Exchange (PEX). Once paired, peers stream data directly over encrypted TCP/WebSocket connections.
2. Relay Fallback (TURN-based): When peers reside on restrictive carrier-grade NATs (CGNAT) or corporate networks, CDUS transparently routes encrypted chunks through a lightweight Go relay. Chunks remain end-to-end encrypted; the relay forwards raw bytes without decrypting payloads.