Files
EchoHub/docs/changelog/v0.2.11.md
T

24 lines
3.0 KiB
Markdown

# v0.2.11
EchoHubSpace directory protocol overhaul: authenticated server registration with persistent claim tokens, near-real-time user-count updates, and richer server metadata (tags, multi-host, version). Coordinated cutover with the EchoHubSpace directory deploy.
## New Features
- EchoHubSpace claim-token authentication — the directory issues a per-server claim token on first registration, persisted atomically alongside the SQLite database (chmod 0600 on Unix). Subsequent reconnects authenticate with the token instead of relying on raw hostname-squatting protection. Token survives both client and directory restarts; lost tokens require an admin-side `DELETE /api/servers/{id}` on the directory to recover
- Server tags — public servers can advertise topic tags via the new `Server:Tags` config array, surfacing as filter facets in the EchoHubSpace browser
- Multi-host advertisement — a single server can register multiple hostnames (e.g. apex domain, IPv6, alias domains) by listing them in `Server:PublicHosts`. All hosts route to the same directory row
- Server version sent to directory — the EchoHubSpace browser shows what version each public server is running, pulled from the server's assembly informational version
- Operator-facing `GET /api/server/directory` endpoint (Admin role required) — returns `ServerId`, `IsRegistered`, `LastRegisteredAt`, `LastError`, and any `ConflictingHosts` for support tickets. Never exposes the claim token itself, only a `HasClaimToken` boolean
## Refactoring
- Replace 30s polling with event-driven directory updates — `PresenceTracker` now raises `UserCountChanged` only when the distinct user count actually changes (multi-tab/multi-connection users no longer trigger). `ServerDirectoryService` consumes via a single-slot `Channel<int>` (latest-wins coalesces bursts) with a 1-second min-interval throttle. Directory reflects user-count changes within ~1s instead of up to 30s stale
- Wrap directory hub responses in a `Response<T>` envelope with `IsSuccess`/`Data`/`Errors`/`Version` shape — protocol version is pinned client-side (currently `1.0`); mismatches trigger a permanent-failure stop with operator-facing log
- Stop attempting re-registration after permanent failures (`HostAlreadyClaimed`, `InvalidToken`, `HostConflict`, `InvalidInput`) — the directory no longer terminates the connection on these errors, so the client suppresses re-register on `Reconnected` to avoid tight retry loops. Operator must restart the server after fixing config
## Configuration
- **Breaking**: `Server:PublicHost` (string) renamed to `Server:PublicHosts` (string array). Public servers must update `appsettings.json` — single-host deployments use a one-element array
- New `Server:Tags` (string array) — defaults to empty
- New optional `Server:DirectoryClaimPath` — overrides the path of the persisted claim file. Defaults to a `directory-claim.json` next to the SQLite database. Treat the file as a secret; back it up alongside the database