# Relays that say what they keep

**Status:** Exploring · **Opened:** 2026-07-10

Not one relay or media host we could find publishes a contractual durability promise — free or paid. The measured economics explain why: 95% of free relays can't cover operating costs from zap income, and 64% receive zero zaps at all (Wei & Tyson, CoNEXT '25). Users treat relays as storage; the economics say they're caches. NIP-11 lets a relay describe its operational limits, and NIP-66 monitors whether a relay is *alive* — but there is no standard, machine-readable way for a relay to declare how long it keeps events, and nothing that checks such a declaration against observed behavior.

## The gap

Durability on Nostr is currently a vibe. A user picking write relays, or an organization with records-retention obligations, has no signal to select on — liveness isn't retention, and payment isn't a promise.

## What we're sketching

Three pieces that only work together: a machine-readable retention declaration (a NIP-11 extension or companion event), NIP-66-class monitors that verify claims longitudinally — "promised 30 days; observed 90" — so declarations earn reputations, and client UX that surfaces it at relay-selection time. Declared retention that nobody verifies is marketing; verified retention is infrastructure.

## Why us

We run the archive-relay pattern ourselves (own strfry as source of truth, public relays as reach), and we build for civic organizations that answer to actual records-retention rules.

## Receipts

- [Wei & Tyson — An Empirical Analysis of the Nostr Social Network](https://arxiv.org/abs/2402.05709v2) — 95% / 64% / p90 zap-income figures
- [NIP-11: Relay Information Document](https://github.com/nostr-protocol/nips/blob/master/11.md) — `limitation` fields exist; no retention promise does
- [NIP-66: Relay Discovery and Liveness Monitoring](https://github.com/nostr-protocol/nips/blob/master/66.md) — liveness, not durability