Use protocols, not services

Original Article ↗ Hacker News Discussion ↗

Protocols vs services and the IRC/Discord split

  • Freenode → Libera is cited as evidence that open protocols let communities escape hostile operators: users largely just moved servers.
  • Others argue it was a “last nail in the coffin” for IRC, with many projects moving to Discord; the migration shows that convenience and features (history, media, voice) trump protocol purity for most groups.
  • Several comments stress that people consciously traded freedom and portability for rapid feature development and polished UX; when enshittification hits, it’s important to remember that trade was made.
  • Discord’s success is attributed less to protocol superiority and more to subsidized hosting, integrated voice/video, and piggybacking on existing gaming communities.

XMPP, Matrix, Nostr and protocol design

  • XMPP gets renewed enthusiasm: extensible, standards-oriented, and capable of Discord‑like features; some are building new clients and highlight niche uses like managing network switches.
  • Critics say XMPP introduced centralization and identifier leakage compared to IRC, and that XML is a liability (complex, easy to mis-implement) rather than a strength.
  • Matrix is viewed by some as a good identity model but a difficult, heavyweight protocol; others link to criticisms about its complexity and security.
  • Nostr is praised for simplicity and offline/sneakernet use, but its identity and relay design are called fragile, lossy, and prone to centralization via “sticky” relays.

Identity as the real problem

  • Many commenters think the core issue isn’t protocols vs services but identity vs applications.
  • Losing a Gmail account or domain means losing a de facto identity; people want identities independent of any provider.
  • Proposed directions: custom domains, DIDs, atproto-style schemes, or government-backed digital IDs with strong privacy guarantees.
  • There’s tension between needing sybil resistance (to fight spam/abuse) and distrust of governments or corporations as ultimate identity authorities.
  • Views diverge on whether identity should be durable or deliberately easy to discard; durability is useful for accounts and reputation, but also increases risk after compromise.

Government control and regulation

  • The article’s claim that you “can’t” enforce age verification or similar rules on decentralized protocols is contested.
  • Skeptics argue states can simply pressure DNS, payment processors, and datacenters, or punish a subset of operators until the rest comply.
  • Supporters counter that attacking thousands of small nodes across jurisdictions is much harder than regulating a few large platforms, though most admit this advantage is “for now” and politically contingent.

AI, spam, and economic shifts

  • Some see decentralized protocols as a counterweight to AI-fueled centralization: LLMs make app-building cheap, so competition rises and margins fall, which may favor small, protocol-friendly tools.
  • Others worry about “100 billion bots” overwhelming any open protocol with spam, scams, and manipulation; cost may not be a limiting factor if attackers profit.
  • Ideas floated: stricter gatekeeping, phone-number–based trust scores, layered networks with increasing authenticity, or stronger identity systems. No consensus on a clean solution.

Usability and self‑hosting hurdles

  • Several people report painful experiences self-hosting Matrix or XMPP: complex setups, multiple components (TURN, Livekit), flaky NAT traversal, poor video quality.
  • Nostr is seen as philosophically appealing but immature: limited clients, rough UX, incomplete long-form and email-like use cases.
  • This leads to a recurring theme: protocols without high-quality, easy-to-deploy reference implementations remain toys for enthusiasts rather than mainstream replacements for Discord/Slack.

Examples of protocol-centric systems

  • Plan 9’s 9P-based “gridchat” is mentioned as a pure “protocols, not services” environment: chat, media, and editing all wired together via a shared filesystem-like protocol and small scripts, giving users total client-side control.
  • Local-first and peer-to-peer designs are highlighted as the next frontier that could push services further out of the loop, though they still face the same identity and spam challenges.