Parley is a federated chat system built on a simple bet: the infrastructure for decentralised communication already exists in DNS, HTTPS, and IRC — nobody needs to invent new client software or new identity schemes. Users get email-style addresses (alice@foo.com), instances discover each other through SRV records and well-known JSON documents, messages travel as signed JSON over HTTPS, and the entire network presents itself to ordinary IRC clients like irssi, WeeChat, or Textual. No plugins, no special apps, no accounts on someone else's platform. The federation model is deliberately minimal. When Bob at bar.com messages Alice at foo.com, bar.com looks up foo.com's SRV record, fetches its ed25519 public key from a well-known endpoint, signs the message, and POSTs it to foo.com's inbox. Both instances then exchange peer lists via gossip, and a mesh forms without manual configuration. The identity layer mirrors Salty IM's approach: DNS plus well-known identity documents, nothing proprietary. Channel architecture splits into two types. Global channels (prefixed with #) replicate across every linked instance with members present — nobody owns them, so there are no operators and no topic control. Local channels (prefixed with &) never leave the instance. Moderation works through block lists rather than channel ownership: ban a user by mask, ban an entire instance by wildcard, and admins can enforce instance-wide blocks. This is a deliberate design choice that trades traditional IRC power structures for federation-compatible access control. The system handles several problems that have historically plagued federated chat. Scrollback follows the account, not the client — CHATHISTORY pages through channels and DMs, read markers sync across devices, and history survives instance downtime via SQLite with full-text search. Authentication supports passwords, IRC tokens, OpenID Connect, and reverse-proxy identity headers. Accounts are managed at runtime through a CLI tool, admin page, or HTTP API, not baked into config files. Parley is currently a working proof of concept, not production-hardened software. The README is transparent about this. But the design document is unusually complete for a proof of concept: it specifies the wire protocol, the discovery mechanism, the configuration surface (flags and environment variables only, no config file), and ships with a Docker Compose demo that stands up CoreDNS, a local CA, and two federated instances. The configuration philosophy — flags for startup-time decisions, database settings for runtime changes — suggests someone who has operated real infrastructure. The structural question is whether Parley can succeed where XMPP federation slowly died and Matrix federation remains complex. Parley's advantage is radical simplicity: no new client protocol, no new client software, no room state resolution algorithm, no DAG. Its disadvantage is the same — IRC semantics impose hard limits on what federation can express. Global channels with no ownership, no topic, and no operators will feel liberating to some communities and ungovernable to others. The extractive dimension here is essentially zero. This is open-source infrastructure that runs on your hardware, uses your domain, stores data in your SQLite database, and connects to clients you already own. The value proposition is entirely generative: it creates new communication capability without requiring anyone to rent access from a platform. Whether the proof of concept hardens into something production-grade is the only open question.
