waste-rs/FUTURE.md

waste-rs — Project Roadmap

A modern reimplementation of WASTE: decentralized, friend-to-friend encrypted mesh networking with chat and file sharing.

---

Design Philosophy

WASTE's core idea is still sound: small trusted groups, no central server, encrypted everything, equal nodes. This project preserves that philosophy while replacing the aging crypto and adding proper NAT traversal.

---

Architecture

Core Daemon — Rust

The networking core (mesh routing, encryption, peer management, file transfer) is implemented in Rust because:

• Memory safety without GC — critical for a long-running daemon handling untrusted peer data
• tokio — excellent async I/O for managing many concurrent peer connections
• ring — modern, audited cryptography (ChaCha20-Poly1305, Ed25519, X25519)
• Native cross-compilation — Windows, Linux, macOS, and eventually Android/iOS

UI Layer — TUI first, Tauri later

Step 1: Terminal UI (ratatui + crossterm)

The daemon exposes everything over the IPC socket, so a TUI is purely a client — it connects to port 17337, renders IpcEvent lines, and sends IpcCommand lines back. No mesh or crypto knowledge required.

• ~400–600 lines of Rust, two crate dependencies
• Works on Windows, Linux, macOS, and over SSH
• Validates the IPC API design before it gets frozen
• Fast to iterate: restart the TUI, daemon keeps running

This is the right first UI milestone. It gives you a real working chat interface without any build toolchain beyond cargo.

Step 2: Tauri GUI

Once the IPC contract is stable, wrap it in a proper desktop app:

• Rust backend shares code directly with the daemon
• Web frontend (React/TypeScript) for a modern chat UI
• Ships as a native binary using the OS webview — not Electron
• Works on Windows, Linux, macOS; Tauri 2.x adds Android/iOS

This avoids the wxWidgets ugliness of the original WASTE and the Qt licensing headaches of the VIA fork.

---

Crypto Modernization

Replaces WASTE's original Blowfish/PCBC (broken cipher mode) + RSA with:

Purpose	Algorithm	Replaces
Identity	Ed25519	Unregistered nicknames
Key exchange	X25519 ECDH	RSA
Symmetric encryption	ChaCha20-Poly1305	Blowfish/PCBC
File integrity	SHA-256	CRC

---

Protocol Modernization

NAT Traversal

WASTE's longstanding problem: one party needs an open port. The fix:

1. ICE-style UDP hole punching (STUN) — try direct connection first
2. Relay fallback — a lightweight relay server (already implemented) forwards encrypted blobs when hole punching fails; the relay never sees plaintext

Transport (future consideration)

Consider migrating from raw TCP to QUIC (via quinn) for:

• Connection multiplexing — multiple streams per peer without head-of-line blocking
• Better NAT behavior out of the box
• Built-in TLS 1.3

Identity & Bootstrapping

• Each peer generates an Ed25519 keypair on first run — the public key is their identity
• Invite via a signed .waste-invite file shared out-of-band (email, Signal, etc.)
• The invite carries: current IP:port hint + public key + short-lived signature
• Once two peers connect, they gossip each other's addresses to mutual friends
• At small group sizes (10–50 nodes) a local known-peers list is sufficient — no DHT needed

---

Roadmap

Near-term

• Invite file format (.waste-invite) for easy peer onboarding
• Relay-assisted message delivery (daemon → relay → peer when no direct path)
• Gossip-driven auto-connect (friends-of-friends)
• Ping/pong keepalive with peer timeout and reconnect

Medium-term

• TUI client (ratatui + crossterm) — chat pane, peer list, text input
• UDP hole-punching (STUN) in nat.rs
• File transfer implementation (FileOffer/FileChunk protocol already defined in proto)
• Message persistence (SQLite via rusqlite)
• Tauri GUI wrapping the daemon (sidebar: peers, rooms, chat history)

Long-term

• QUIC transport via quinn
• Mobile daemon (Android/iOS via Rust cross-compilation)
• Kademlia DHT for larger-scale peer discovery (if the group outgrows static bootstrap)

---

What We Keep from WASTE

• Small, trusted friend groups — not a public network
• No central authority — the relay is dumb and optional
• Encrypted everything — the relay forwards opaque blobs
• Equal nodes — no privileged "servers" among peers
• Simple onboarding — one file or one URL to join a group