Skip to main content

Introduction

Why EphemeralNet?​

Shardian designs infrastructure for teams that take deletion guarantees as seriously as confidentiality. EphemeralNet is our C++20 research daemon for data that must disappear: a P2P protocol that uses enforced time-to-live (TTL) semantics as a first-class constraint. Instead of mutating BitTorrent-style swarms into ad-hoc retention policies, EphemeralNet treats each manifest, chunk, and routing hint as expendable material whose lifetime is negotiated at the edge and cryptographically enforced.

Every node runs eph, a compact daemon that exposes a gRPC-inspired control socket and a remote-style CLI. The kernel pairs a TTL-aware Kademlia DHT with ChaCha20/HMAC sessions, Shamir's Secret Sharing for key pulverization, and deterministic cleanup hooks so expired state evaporates across storage tiers, routing tables, and manifests simultaneously. The outcome is a mesh that can gossip, store, and fetch like a modern CDN overlay—while guaranteeing the data falls out of existence on schedule.

Unlike archival systems, EphemeralNet optimizes for temporal correctness: it verifies that a chunk cannot outlive its deadline, even if peers go offline. That makes it ideal for incident data rooms, temporary collaborative edits, and any workload that needs Zero Trust guardrails for deletion.

High-Level Architecture​

Feature Highlights​

  • TTL-native Kademlia: XOR distance routing with per-entry expiry metadata so announcements self-destruct without operator intervention.
  • E2EE everywhere: ChaCha20-Poly1305 for transport, HMAC-SHA256 for protocol integrity, and rolling session key derivation per peer.
  • Shamir-backed key expiry: decryption keys are sharded across the DHT; when shards expire, the data becomes irrecoverable even if chunks linger in caches.
  • Control-plane proof-of-work: adaptive PoW envelopes handshake, store, and manifest operations to blunt Sybil abuse while keeping honest peers fast.
  • NAT traversal + relay fallback: built-in STUN/TURN learning, auto-promoted public endpoints, and an event-driven relay server for high-fanout bursts.
  • Deterministic storage hygiene: in-memory and optional disk tiers, secure wiping, TTL audits, and notification hooks that synchronize cleanup with manifest withdrawal.
  • Observability first: Prometheus exporters for TTL drift, PoW pressure, DHT saturation, and per-peer health to keep SREs ahead of runaway swarms.
  • CLI/daemon parity: the same eph binary exposes both the background service and a remote control surface for automation-friendly operations.