16 Commits
Author SHA1 Message Date
thepraandClaude Opus 5.5 5f56681c01 Everything on, phase 1: geolocation fetches itself, the deploy signs in as @thepra, the crawler is on, sign-up by invitation
Build / Build (push) Successful in 5m1s
Deploy / privapub.thepra.dev (push) Successful in 5m39s
Owner decisions (2026-10-04, recorded in docs/ROADMAP.md): production runs everything that is built, and nothing waits
on a person running a command.

- Geolocation updates itself. GeoUpdater, a hosted service, checks daily whether each DB-IP Lite database was built this
  month. If not, it fetches this month's, or last month's early in the month. It installs a file only once it opens as
  the right kind of database, then swaps it in atomically, and the locator reloads at once. Lookups now run under the
  lock, so a reload can no longer dispose a reader mid-lookup. The systemd timer, its script and their setup.sh lines
  are gone: the root step they needed never happened, and none is needed now. /stargazing names the database in use.
- The admin CLI runs after the app is built, with every service and nothing started.
  - `create-root <login> [--admin]` takes the password on stdin; it is how the first login is made while sign-up is
    closed.
  - `smoke <persona>` keeps the root `deploy-smoke` and an undiscoverable persona, and gives the root a new password
    on every run.
- The deploy signs in as @thepra. It runs the CLI, gets a token through the real OAuth flow (tools/smoke/oauth.sh,
  moved out of the pasture's privapub_token, which now uses it), checks the signed-in API and that @thepra is
  undiscoverable, then revokes the token. PRIVAPUB_SMOKE_TOKEN is gone.
- The deploy also fails when:
  - NodeInfo and the instance API disagree about registrations;
  - /stargazing does not say the crawler is on;
  - the geolocation databases are missing or more than 40 days old.
- The crawler is on in production, seeded with ten large servers of different kinds. FEDERATION.md now describes it
  and how to opt out.
- One registrations switch (Registrations:Mode, default Invitations; Open in tests and the pasture). It is read by
  open sign-up (403 when closed), NodeInfo `openRegistrations`, and v1 and v2 of the instance API, so they can no longer
  disagree. Before, NodeInfo said open and the instance API said closed. Group invitations always work, so
  invites_enabled is true.
- A persona edit through /clientapi no longer resets what the Mastodon API set (discoverable, locked, quote policy…):
  the theme is merged into the settings instead of replacing them.

650 tests pass. The deploy's smoke step was rehearsed against the pasture's PrivaPub.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01ELjqpznMFMNrJoJUj6K5p2
2026-10-04 02:37:38 +02:00
thepraandClaude Opus 5.5 2645dea26f T8: inbox gaps, jobs, migrations and pages; a deleted remote account's posts are hidden
Owner decision (2026-10-03, "A remote account deletes itself"): its posts are kept but
hidden everywhere.
- Post.AuthorGone (additive bool). DeleteHandler's actor-delete branch sets it on every
  post whose ActorURI is the actor (one update-many), besides dropping its follows and
  timeline rows as before. RemotePosts.Build sets it on a post stored later for an
  account already marked Deleted.
- One rule in VisibilityPolicy: IsShown (not deleted, author not gone), IsPublic and
  CanSee exclude AuthorGone, plus Shown(post) for loaded posts.
- Lookups by id answer 404 through CanSee (statuses/:id and every sub-route, context,
  bookmarks, favourites, polls, reactions, search); provenance, account statuses,
  home/public/tag timelines, notifications, conversations, reblogged_by, the clientapi
  home and post/DM lists, a community's outbox and our Announces filter on IsShown or
  IsPublic; the Mastodon mapper never renders a hidden post or a boost of one.

Tests (30 new):
- AuthorGoneTests: the rule, the handler (posts kept, boosts included, follows and rows
  gone), a post fetched after the delete, and 20 Mastodon/ActivityPub lookups over HTTP
  seen before and hidden after.
- InboxGapTests: actor Update refresh (name, sanitised summary, key rotation in place
  and to a new key id) even with an older `updated`; Undo{Follow} by activity id and by
  object; Reject of our QuoteRequest (and a stranger's ignored); group-wrapped
  Announce{Like} and Announce{Undo{Like}}; a locked persona's pending follow,
  FollowRequest notification, and Decide accepting and rejecting with the original Follow.
- JobHandlerTests: AncestorsJobHandler up to its depth limit; PollRefreshJob and
  PollCloseJob (local and remote polls); InstanceDescriber from a peer's NodeInfo and
  the weekly dedupe through ObjectRecords; LinkPreviews for public posts only;
  DeliveryJobHandler outcomes (2xx, 404/410, 429/503 with Retry-After in seconds and as
  a date, 5xx) and a signature and Digest the peer can verify; MediaJanitor.Sweep;
  OAuthPruner.Prune.
- MigrationTests: _003, _004, _006 and _007 on seeded rows.
- PublicPagesTests: /@user and /@user/{id} (visibility, junk ids, exact CSP,
  Referrer-Policy and nosniff), circle 404, community page, the instance actor,
  ActivityPub redirects, and markup escaped in posts, titles and bios.

Production changes besides the rule:
- LinkPreviews.Handle re-checks that a post is still shown and public (the rule
  Wanted applies) before fetching anything; before, only enqueueing checked it.
- The legacy /clientapi post and DM lists no longer return soft-deleted posts.
- MediaJanitor.Sweep and OAuthPruner.Prune are the loop bodies, now public and tested.
- InstanceDescriber.Address: a protected virtual identity seam so a test can point
  the https NodeInfo addresses at a plain-http peer; production behaviour unchanged.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01ELjqpznMFMNrJoJUj6K5p2
2026-10-03 11:53:23 +02:00
thepraandClaude Opus 5.5 c5e4934ba6 T5: OAuth and the client API over HTTP
96 integration tests through PrivaPubHost, the real pipeline end to end:
- OAuth: the token's subject is the persona, and neither the token response,
  verify_credentials nor the stored token entries name the root. A wrong password shows
  an error and sets no login cookie; a login without the antiforgery token is a 400; the
  return address never leaves the site; deny answers access_denied with no code; another
  root's persona re-renders the choice with no code; a banned root is sent back to the
  login and a code issued before the ban buys no token; force_login asks again; a code
  works once and its reuse revokes the token it bought; password and refresh_token
  grants are refused; a client_credentials token gets 401 on user routes; a read-only
  token gets 403 with a Mastodon error on POST /api/v1/statuses; follow covers
  read:follows; revoke works; the login and authorize pages send their CSP and no-store;
  the 11th /oauth/login from one address in a minute is a 429.
- Accounts: sign-up, duplicates in any case, invalid models answer 400 with a message,
  login and logout, recovery email, settings, password change, invitation sign-up and
  login (refusing a persona named after the login), recovery without an email, through
  an unreachable mail server, with a wrong and with a valid code, token refresh, the 11th
  sign-up from one address, expired, garbage and foreign-key JWTs.
- Personas: a rootId in the body is ignored, the username regex and reserved names hold,
  personas and groups share ReservedName, an update delivers Update{Person} to followers,
  PublishedOn and the id's day fall within two weeks before creation, the list holds only
  one's own personas, another root's persona cannot be updated.
- Groups: communities and circles are created, joining takes the code and the password,
  members leave and owners cannot, a remote follow request becomes a member only on
  approval, a circle never shows in lookup, account by id, v2 search or /@name, and its
  /flock and /wardens answer 404 unsigned and to non-members, 200 to a member's signed GET.
- Moderation (Exclusive, it suspends localhost): ban, unban and remove; non-admins get
  403; reports are listed without the reporter and resolved; domain blocks are inserted,
  listed and deleted, bad domains refused, and a suspended server's delivery is answered
  202 and kept nowhere; the data endpoints.
- AdminCommands: exit codes 0, 1 and 2 and the resulting policies.

Fixed:
- A banned or removed root kept using /clientapi with its JWT until it expired: only /api
  re-checked the root. JwtEvents.TokenValidated now loads the root and fails the request
  when it is banned or deleted, and takes the policy claims from the database, so a
  demoted admin loses admin at once (and a promoted one gains it).
- The 401 and 403 bodies JwtEvents writes were PascalCase while every other /clientapi
  answer is camelCase; they now use the web defaults.
- /clientapi/user/sniff/again (token refresh) answered an empty 200; it now answers a
  fresh JwtUser, like login.
- Password recovery answered SMTP reply codes as HTTP statuses (421, 454, 554, and 550 for
  an invalid address); a mail server failure is now 503 and an invalid address 400.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01ELjqpznMFMNrJoJUj6K5p2
2026-10-03 11:53:23 +02:00
thepraandClaude Opus 5.5 cee85309b8 M9 (first part): the means to locate a server
- IConnectedAddresses remembers which address each host's last connection reached, set in
  SafeHttpHandlerFactory's connect callback. That is once per pooled connection, with no
  second DNS lookup, and the address is never stored.
- IGeoLocator / DbIpLocator reads the offline DB-IP Lite city and ASN databases (MaxMind
  .mmdb, via MaxMind.Db). It maps memory, swaps to new files within ten minutes, rounds
  coordinates to one decimal, never looks up a private address, and answers nothing
  when the files are missing. CdnNetworks names the CDNs whose edge addresses say nothing
  about where a server is.
- deploy/max/geo-update.sh fetches this or last month's databases, checks them and swaps
  them in atomically. The privapub-geo timer runs it monthly as www-data, and setup.sh
  installs the directory, the script, the units and a first download.

Describing servers will use these in M8.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01ELjqpznMFMNrJoJUj6K5p2
2026-10-03 11:33:30 +02:00
thepraandClaude Opus 5.5 90a7e38ce9 M6: provenance fixes, stored features, community edits that are real edits
Build / Build (push) Successful in 2m34s
Deploy / privapub.thepra.dev (push) Successful in 2m53s
- A fetched ObjectRecord no longer wears the signature, key, signed headers and @context
  of the activity that caused the fetch, and its ReceivedAt is the fetch's own time. Its
  activity fields now name the trigger. Provenance answers signature null and
  fetchedBy "instance-actor". Migration _008 clears the old fetched records.
- ObjectRecords store their ObjectFeatures extensions and context namespaces (migration
  _009 fills older records in batches), so statistics can count them. Provenance reads the
  stored lists.
- ForeignAvatar.Features records what an actor document uses (ActorFeatures): featured,
  shared inbox, FEP-521a, FEP-8b32, identity proofs, Misskey fields, indexable,
  undiscoverable, locked, key size, and more.
- RemoteEdits.Apply and RemoteDeletes.Remove are shared by the Update, Delete and Announce
  handlers. A community-wrapped Update is now an edit only when newer, refreshes polls
  and media otherwise, revises the ObjectRecord and re-resolves quotes. A
  community-wrapped Delete now tombstones the object, removes its ObjectRecord, reblogs
  and timeline rows, and lowers the reply count. Any deleted quoting post lowers the
  quoted post's quote count.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01ELjqpznMFMNrJoJUj6K5p2
2026-10-03 11:15:21 +02:00
thepraandClaude Opus 5.5 00b2685cf4 T1: tests stop sharing state they don't own
- JobQueue takes an optional scope, so a test's worker leases and reaps only its own jobs.
- The dead-host delivery test runs alone (Exclusive), on its own jobs, and cleans up the
  breaker rows it trips; the breaker has tests of its own on unique hosts.
- Index and migration tests run alone: they drop indexes and rewrite every post.
- DomainBlocks.Load replaces reflection and a database-wide block in tests.
- Harness.Outgoing sees only deliveries queued since the harness started: Peer ports are
  reused within a run, which made the circle test flaky.
- Two pure-logic tests leave Mongo-gated classes, so CI runs them.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01ELjqpznMFMNrJoJUj6K5p2
2026-10-03 10:36:15 +02:00
thepraandClaude Opus 5.5 bb9bd71391 P5, first batch: summaries stop hiding articles, personas get reachable names, blocks federate
Build / Build (push) Successful in 50s
Deploy / privapub.thepra.dev (push) Successful in 1m4s
- A remote `summary` is a content warning only on a Note or Question, or when `sensitive` is set. On Articles,
  Events, Videos, Pages and Audio it is an excerpt (WordPress teasers, Mobilizon dates and places, Mbin titles) and
  is now kept as `Post.Excerpt` instead of hiding the post. `Post.ObjectType` records the remote type, and the
  Mastodon API shows a remote non-Note object's title above its body again.
- Persona usernames must match `^[a-z0-9_]+$`, as groups already did; a name outside it was unreachable from
  Mastodon and Misskey.
- `postingRestrictedToMods` is defined in our JSON-LD context (Iceshrimp.NET drops undefined terms).
- `Vary: Accept` on actor and object URLs.
- Owner decision: blocks federate. A block sends `Block`, an unblock `Undo{Block}`; checked live against GoToSocial.
- Owner decision: a persona's and a group's `published`, and the day in new ids, is a random day up to two weeks
  before creation, so personas made the same day no longer share a date. Migration _007 gives existing ones theirs.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012CzABvBkbcFqoHdmi8b9WB
2026-10-01 17:39:58 +02:00
thepraandClaude Opus 5.5 31542a181e Remote posts take their timeline place on arrival, and an unedited one has no edit date
Found by the first live run against GoToSocial:

- A remote post's id came from its published second plus random bytes, so a reply arriving in the same second as
  the post it answers could sort under it, and a late arrival landed behind a client's since_id. A post published
  within the last hour now gets an id for its arrival; backfill keeps its published time.
- NoteParser.Time returned default(DateTime) for a missing "updated", so every remote post was stored as edited in
  year one. Migration _006 clears the stored ones.
- published now carries milliseconds, so peers that derive ids from it (GoToSocial, Mastodon) keep our posts in
  order within a second.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012CzABvBkbcFqoHdmi8b9WB
2026-10-01 13:19:16 +02:00
thepraandClaude Opus 5.5 70b4c45eb0 The dead-host test tolerates a loaded machine
It asserted fewer than ten requests to the dead host by the time the live
one arrived, which a busy workstation can exceed without the per-host
limit being broken; the live delivery arriving at all is the point.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012CzABvBkbcFqoHdmi8b9WB
2026-10-01 12:07:44 +02:00
thepraandClaude Opus 5.5 5a13597c00 Persona and group ids say which day, nothing more
An ObjectId carries its creation second and a per-process counter, so two
avatars made one after the other by the same login got ids a few counts
apart: a link between personas that every Mastodon client would have
seen. Avatar and Group now generate ids from the UTC day plus eight random
bytes (still valid ObjectIds), and a remote post's id is made from its
published time with a random tail, so posts page in the order they were
written rather than the order they arrived.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012CzABvBkbcFqoHdmi8b9WB
2026-10-01 11:25:48 +02:00
thepraandClaude Opus 5.5 e6a362c0b8 Domain blocks: suspend, silence, reject media
DomainBlock (domain, severity, reject-media, public and private comment)
covers the domain and its subdomains. Admins manage them under
/clientapi/admin/domainblocks/{list,insert,delete}; the set is kept in
memory, reloaded on every change and at most five minutes stale.

A suspended domain is refused by FederationHttp.IsAllowed, so nothing is
fetched from it and no job delivers to it, and the inbox drops its
activities with a 202 before fetching any key. Reject-media strips the
attachments of posts from that domain. Silence is recorded for the
timelines and notifications that arrive in P1.2.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012CzABvBkbcFqoHdmi8b9WB
2026-10-01 11:25:05 +02:00
thepraandClaude Opus 5.5 e9529310fe One Post model for every visibility, read from notes the way peers write them
Post gains what federation and the Mastodon API need: Visibility (Public,
Unlisted, FollowersOnly, Direct, Circle, LocalGeo), the author's account
id, to/cc, the Create's id, url, context, quote, InReplyToURI and the
parent's author, a separate SpoilerText next to the title, language,
mentions, hashtags, remote attachments (alt text, blurhash, focus, size),
reply/favourite/reblog counters, revisions, EditedAt and DeletedAt.

Direct messages are Posts with Visibility Direct and a ConversationId;
migration _005 copies DmPost rows across with their ids and fills the new
fields of existing posts. DmPost is left in place so a rollback still sees
the old messages.

Inbound:
- NoteParser reads Note, Article, Page, Question and media types: content,
  then contentMap, then _misskey_content; summary as the spoiler and name
  as the title; Mention and Hashtag tags; attachments; a PeerTube-style
  list attribution prefers the person over the channel; quote URIs.
- Addressing classifies like Mastodon, finding followers-only by the
  author's own followers URL (now stored on ForeignAvatar), not by a
  "/followers" suffix.
- Create keeps a post when a local persona is addressed or mentioned, when
  it replies to a local post (the parent's reply count goes up) or when a
  community it follows is addressed; an unsolicited public post is not
  stored. Update keeps the previous version as a revision.

The outbox and object endpoints serve only Public and Unlisted posts.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012CzABvBkbcFqoHdmi8b9WB
2026-10-01 11:17:14 +02:00
thepraandClaude Opus 5.5 d1a91c40c4 Deliveries run on a Mongo job queue with leases, backoff and per-host limits
The single serial DeliveryWorker is replaced by Infrastructure/Jobs:
- Job rows are leased with one FindOneAndUpdate (oldest RunAt first, a
  two-minute lease) and a reaper returns expired leases every 30 s;
- enqueueing wakes the workers, which otherwise poll every five seconds;
- delivery runs eight at a time with at most two per host, so a slow or
  dead server holds two slots, not the queue;
- a failure waits n^4 + 15 + jitter seconds (Mastodon's curve) for up to
  16 attempts; a 4xx other than 408/429 is final, a 429 honours
  Retry-After;
- RemoteInstance is a per-host circuit breaker: ten consecutive failures
  quarantine a host for an hour, doubling to a week, and its jobs wait
  without spending attempts;
- a delivery is queued once per activity and inbox (unique DedupeKey), and
  finished jobs expire after seven days (TTL on FinishedAt).

Migration _004 moves pending Delivery rows into jobs and marks them
abandoned, so a rollback to the old worker cannot send them twice.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012CzABvBkbcFqoHdmi8b9WB
2026-10-01 11:09:35 +02:00
thepraandClaude Opus 5.5 d2f0c7a14e Unique indexes, one username space, and entity maps warmed before use
Infrastructure/Data/Indexes runs at every start, after the migrations:
unique on Post/DmPost ObjectURI, ForeignAvatar ActorURI, the Follower
triple, RootToAvatar, RootUser UserName and ReservedName; plain indexes on
the lookups the services actually make (PublicKeyId, author and group
post listings, ParticipantsKey, the delivery queue).

Migration _001 runs first and removes the duplicates the races could
already have left (keeping the newest actor row, the oldest post, the
accepted follower), then fills ReservedName from every avatar and group.

ReservedName is one username space for personas, groups and the instance:
a name is reserved by an insert the unique index arbitrates, before the
avatar or group is saved, so two simultaneous sign-ups cannot both get it.
A short list of names (admin, support, abuse, postmaster, ...) is never
available.

EntityMaps.Warm touches every entity's collection one at a time before
anything else runs. MongoDB.Entities maps the Entity base class on first
touch, and two types mapped at once throw "An item with the same key has
already been added" and stay broken for the life of the process; the
parallel test run hit it, and the delivery worker racing a request could
have too.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012CzABvBkbcFqoHdmi8b9WB
2026-10-01 11:01:08 +02:00
thepraandClaude Opus 5.5 bf7c88ce71 Remote HTML is sanitized before it is stored
S5 of the roadmap. ContentSanitizer wraps HtmlSanitizer with Mastodon's
allowlist: the inline and list tags Mastodon keeps, href/rel/class and
the list attributes, microformat and mention/hashtag/ellipsis/invisible
classes, Mastodon's link schemes, every link rel=nofollow noopener
noreferrer, relative links unlinked, headings folded to a bold paragraph,
and the contents of script, style, svg, iframe and friends dropped rather
than kept as text.

Post and DmPost gain ContentHtml (what is shown) and ContentFormat
(Markdown for local, Html for remote). Inbound Create and Update, and a
remote actor's biography, are sanitized on the way in; local posts store
their Markdig rendering. Migration _002 does the same to what is already
stored, and migrations now run at startup.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012CzABvBkbcFqoHdmi8b9WB
2026-10-01 10:54:17 +02:00
thepraandClaude Opus 5.5 ccc3597699 Every outbound federation request goes through one guarded client
Infrastructure/Http adds the client the roadmap's S3 and S4 ask for:
- the connect callback resolves the name itself and refuses loopback,
  private, link-local, CGNAT, documentation, multicast, ULA, NAT64, 6to4,
  Teredo and IPv4-mapped/compatible forms, then connects to the vetted
  address, so DNS rebinding cannot swap it afterwards;
- redirects are followed by hand, at most three, each one re-checked;
- bodies are capped at 1 MB after decompression, only JSON media types are
  read, every request has a 15 s budget, and a refused URL is not asked
  again for five minutes.

Actor and WebFinger fetches and inbox deliveries all use it. Test networks
can switch on Federation:AllowPrivateNetworks/AllowPlainHttp; startup
refuses both in Production.

PrivaPub.Tests (xUnit v3) starts with the address table and the fetcher's
limits against an in-process peer; build.yml and deploy.yml run it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012CzABvBkbcFqoHdmi8b9WB
2026-10-01 10:46:09 +02:00