trails/apps/journal/app/lib/federation-delivery.server.ts
Ullrich Schäfer 20be961177 fix(journal): retract federated activities only on public→non-public
A received Delete is recorded by Mastodon as a permanent tombstone —
later Creates for the same URI are silently refused forever. The old
code sent Delete on every non-public save 'just in case' (the comment
even called over-sending harmless); on the 2026-06-07 soak this
tombstoned an unlisted activity's URI before its first real publish,
making its later flip to public invisible on the remote with no error
anywhere.

- visibilityTransitionAction(previous, next): Create on any transition
  to public (re-publish doubles as back-delivery; remotes dedupe by
  id), Delete only when leaving public, nothing for
  non-public→non-public.
- updateActivityVisibility reads the previous visibility and acts on
  the transition.
- design.md documents the tombstone permanence + the user-facing
  consequence (un-publish then re-publish won't resurrect the post on
  remotes that processed the retraction — same as Mastodon's own
  delete-and-redraft).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-07 08:52:23 +02:00

141 lines
4.9 KiB
TypeScript

// Push delivery of local public activities to accepted remote
// followers (spec: social-federation, "Push delivery on local activity
// create"). Enqueue side lives here; the actual signed POST happens in
// jobs/deliver-activity.ts.
import { and, eq, isNotNull } from "drizzle-orm";
import { follows, remoteActors, users } from "@trails-cool/db/schema/journal";
import { getDb } from "./db.ts";
import { federationEnabled } from "./federation.server.ts";
import { enqueueOptional } from "./boss.server.ts";
import { activityObjectIri } from "./federation-objects.server.ts";
export type DeliveryAction = "create" | "delete";
/**
* Which federation action (if any) a visibility change warrants.
*
* The asymmetry matters: Mastodon records a `Delete` as a permanent
* tombstone — a later `Create` for the same URI is silently refused
* forever. So a retraction must only go out when remotes could
* actually have the object (it was public), never "just in case"
* (learned on the 2026-06-07 staging soak, where an unlisted
* activity's no-op save tombstoned its URI before its first real
* publish).
*
* - → public: push Create. Also for public→public — re-pushing is
* harmless (remotes dedupe by id) and doubles as back-delivery.
* - public → non-public: push Delete(Tombstone) — remotes saw it.
* - non-public → non-public: nothing — remotes never had it, and a
* Delete would poison the URI for any future publish.
*/
export function visibilityTransitionAction(
previous: string,
next: string,
): DeliveryAction | null {
if (next === "public") return "create";
if (previous === "public") return "delete";
return null;
}
export interface DeliveryPayload {
action: DeliveryAction;
/** Present for `create` — the job re-reads the row at delivery time. */
activityId?: string;
/** Object IRI; for `delete` this is all that's left of the activity. */
objectIri: string;
ownerUsername: string;
recipientActorIri: string;
}
/** Accepted remote followers of a local user (the delivery audience). */
export async function listAcceptedRemoteFollowers(ownerId: string): Promise<string[]> {
const db = getDb();
const rows = await db
.select({ actorIri: follows.followerActorIri })
.from(follows)
.where(
and(
eq(follows.followedUserId, ownerId),
isNotNull(follows.followerActorIri),
isNotNull(follows.acceptedAt),
),
);
return rows.map((r) => r.actorIri).filter((iri): iri is string => iri !== null);
}
/**
* Fan out one delivery job per accepted remote follower (spec 5.3: a
* job per follower, each with its own retry/backoff lifecycle). No-op
* when federation is off or the user has no remote followers — the
* common case stays a single SELECT.
*/
export async function enqueueActivityDeliveries(
ownerId: string,
activityId: string,
action: DeliveryAction,
): Promise<void> {
if (!federationEnabled()) return;
const recipients = await listAcceptedRemoteFollowers(ownerId);
if (recipients.length === 0) return;
const db = getDb();
const [owner] = await db
.select({ username: users.username, profileVisibility: users.profileVisibility })
.from(users)
.where(eq(users.id, ownerId))
.limit(1);
// Private profiles don't federate — suppress push delivery entirely
// (spec 9.3: flipping to private stops federation).
if (!owner || owner.profileVisibility !== "public") return;
for (const recipientActorIri of recipients) {
const payload: DeliveryPayload = {
action,
activityId: action === "create" ? activityId : undefined,
objectIri: activityObjectIri(activityId),
ownerUsername: owner.username,
recipientActorIri,
};
await enqueueOptional(
"deliver-activity",
payload,
{ source: "enqueueActivityDeliveries", action },
// Spec 5.4: exponential backoff on failure, bounded retry budget.
// 8 retries with backoff spans roughly a day before permanent fail.
{ retryLimit: 8, retryBackoff: true },
);
}
}
export interface CachedRemoteActor {
actorIri: string;
inboxUrl: string | null;
}
/** Read the remote actor cache; null when we've never seen the actor. */
export async function getCachedRemoteActor(actorIri: string): Promise<CachedRemoteActor | null> {
const db = getDb();
const [row] = await db
.select({ actorIri: remoteActors.actorIri, inboxUrl: remoteActors.inboxUrl })
.from(remoteActors)
.where(eq(remoteActors.actorIri, actorIri))
.limit(1);
return row ?? null;
}
/** Upsert the fields delivery learns about a remote actor. */
export async function upsertRemoteActor(fields: {
actorIri: string;
inboxUrl?: string | null;
displayName?: string | null;
username?: string | null;
domain?: string | null;
}): Promise<void> {
const db = getDb();
const { actorIri, ...rest } = fields;
await db
.insert(remoteActors)
.values({ actorIri, ...rest })
.onConflictDoUpdate({ target: remoteActors.actorIri, set: rest });
}