STATS reports the live counts OperServ holds — AKILL and SQLINE bans by
kind, and services ignores — as a quick at-a-glance check. Read-only, so
any operator can run it.
IGNORE ADD [+expiry] <mask> [reason] / DEL <mask> / LIST. While a user
matches, the dispatch choke-point drops their service commands silently.
A mask with an @ matches nick!*@host (ident isn't tracked); a bare mask
matches the nick; matching is case-insensitive with lazy expiry. Admin-
only, and an operator is never ignored (so staff can't lock themselves
out). Node-local and in-memory, like the auth and request throttles — a
fast transient moderation tool, not persisted or federated.
Two channel operator tools:
- MODE <#chan> <modes> [params] forces a channel mode change (TS 1, so it
applies regardless of the current TS). A status-mode target may be given
as a nick — it's resolved to the uid the ircd's FMODE expects.
- KICK <#chan> <nick> [reason] removes a user, sourced from OperServ and
attributed to the operator.
Both admin-gated. To resolve status-mode params, channel-mode parameter
arity became one shared helper (chanmode_takes_param + STATUS_MODES in the
api): the ircd module now delegates to it instead of keeping its own copy,
so burst-parsing and MODE-building can never drift.
Generalise the network-ban store to carry the ircd X-line kind (defaulted
to the user@host G-line for existing records) so one event-sourced,
lazily-expiring, burst-reasserted list backs every ban type, and drive
ADD/DEL/LIST from one implementation per kind:
- AKILL stays the user@host G-line; SQLINE is a nick Q-line, its mask
validated as a nick glob (an @ is refused so an AKILL-shaped mask can't
become an over-broad nick ban). Both are keyed by (kind, mask).
- GLOBAL <message> announces to every user via a $* server-glob notice.
- KILL <nick> [reason] disconnects a user, attributed to the operator.
All four are admin-gated, and OperServ itself stays invisible to non-opers.
Bans and kills go out as typed actions (AddLine/DelLine/KillUser) the ircd
link serialises to ADDLINE/DELLINE/KILL.
When [expire] warn_days is set and email is configured, the expiry sweep
now emails a heads-up to owners entering the warning window (inactive past
the threshold minus the lead time, but not yet expired) for whom an
address is on file — the founder's address for a channel. It's a chance to
keep the record by simply using it before the deadline.
Each idle spell warns at most once: a new AccountExpiryWarned /
ChannelExpiryWarned event marks the record, and the next activity stamp
clears the mark so a later idle spell warns afresh. The warning email is a
first-class api template (fedserv_api:📧:expiry_warning), alongside
the reset and confirm mails.
The first operator module. AKILL manages network bans as a typed,
event-sourced list with lazy expiry, mirroring the patterns the other
modules use:
- AKILL ADD [+expiry] <user@host> <reason> stores the ban and drives an
ircd G-line (applied to matching users already online and to future
connections); AKILL DEL <mask|number> lifts it; AKILL LIST [pattern]
shows the live list, numbered. Admin-only.
- Bans are Global events, so every node holds the list and re-asserts each
one at burst with its remaining duration. Expired bans are hidden lazily
and dropped at compaction.
- Masks are normalised to user@host (a nick! prefix is stripped, both
sides lowercased) and an all-wildcard mask is refused.
Email (send_email) is already reachable from the api via ServiceCtx, and
adds AddLine/DelLine to the action vocabulary for any future X-line kinds.
Accounts and channels now carry a last-activity stamp and are dropped
once inactive past a configured threshold. A login stamps the account, a
join stamps the channel — both coalesced to at most one log entry per day
so routine traffic doesn't bloat the log, and stamps replicate so
activity on any node keeps a record alive network-wide.
Expiry is lazy in the proven style: a periodic pass computes it from the
stored stamps with no per-record timer, reusing the same relation the
vhost and suspension expiries use. The sweep spares operator accounts,
accounts with a live session, occupied channels, and anything an operator
has pinned with the new NOEXPIRE command (NickServ NOEXPIRE <account>,
ChanServ NOEXPIRE <#channel>, both Priv::Admin). Each expiry is announced
to the audit channel and clears the channel's +r.
Configured under [expire] (accounts_days / channels_days); a zero or
omitted field leaves that kind never expiring, and omitting the section
disables expiry entirely.
A [log] channel now receives a one-line announcement of each notable
state change a service command makes — registrations, drops, vhosts,
suspensions, akicks, access and founder changes, bot and oper host-config
changes — naming the actor (by nick, plus account when it differs) and
what changed. The feed reads the typed event log directly: dispatch marks
the log before running a command and diffs it after, so each announcement
is exactly that command's footprint, sourced from the services server so
it reaches the channel without a pseudo-client having to be present.
Private and cosmetic events are deliberately never surfaced: memo bodies,
password/verifier material, greets, auto-join, personal and channel
cosmetic settings. Omitting [log] disables the feed.
- Refuse to start the gRPC accounts API on an empty token: the
constant-time compare treats two empty slices as equal, so a blank
token would let an unauthenticated request through. Mirror the guard
the JSON-RPC endpoint already had.
- Sweep expired votekick/voteban tallies on each vote. A passed vote
removed its own key, but a tally that never reached threshold lingered
forever, so the map could be grown without bound. The now-redundant
per-key TTL reset goes away.
Free-form trailing text (message bodies, kick reasons, topics, metadata)
now passes through a single sanitizing choke-point in the serializer, so
no such text can carry an embedded line break that would smuggle a second
command onto the link. Line-based input closes the obvious path today,
but sanitizing at the point of serialization removes the whole class
regardless of where the text originated (federation, config, future
non-line sources).
- Box the oversized enum variants (Event::AccountRegistered, gossip
Msg::Entry) so small events aren't sized to a full Account; serde is
transparent so the wire/log format is unchanged.
- Use the configured server description in the SERVER line (it was
hardcoded and the config field went unread).
- Derive the ChanServ SET label from the ChanSetting (drops toggle's 8th
arg); rename inspircd from_us -> sourced.
- allow the two framework-idiom lints with rationale (tonic's Status in
authorize; the guest-rename handler's inherent arg count).
- Auto-fixed map_or/contains/split_once style lints.
Canonicalise a requested vhost the way the ircd displays it (a disallowed
underscore becomes a hyphen) before storing and applying, so what we keep
matches what shows on the network. A vhost can no longer be assigned to two
accounts: prepare_vhost normalises, validates, and checks vhost_owner at
every assignment point (SET/REQUEST/ACTIVATE/TAKE/DEFAULT), so two inputs
that collapse to the same host are caught as one.
Approach recommended by siniStar (the ns_nethost normalisation logic).
SET <account> <host> [duration] assigns a vhost that lapses after the
duration (e.g. 30d); a Vhost carries an optional expiry and the store hides
an expired one lazily (like a suspension), so it stops applying on identify/
ON without a sweep. LIST flags temporary vhosts.
A member may request a vhost at most once a minute; a request sooner is
refused with the wait time. In-memory per-account (not event-logged), the
same shape as the identify brute-force throttle.
FORBID <regex> / FORBIDLIST / FORBIDDEL let operators block impersonating
user requests (e.g. (?i)(admin|staff)); REQUEST refuses a matching host.
Matching reuses the linear-time RegexSet, so untrusted patterns are safe.
TEMPLATE <$account...> sets a network auto-vhost pattern; DEFAULT gives a
user $account.users.example with their sanitised account name. HostServ's
node config (offers/forbidden/template) is consolidated into one HostConfig
threaded through the log replay.
A curated self-serve menu: operators OFFER <host> / OFFERDEL <n>, anyone
sees OFFERLIST, and a user picks one with TAKE <n> (assigned and applied
at once). Offers are a global event-sourced list (VhostOfferAdded/Removed,
threaded through apply like bots). SETALL/DELALL alias SET/DEL and GROUP
is a no-op reassurance, since the per-account vhost already covers every
grouped nick.
A vhost may be ident@host, not just host: apply_vhost splits it and sends
CHGIDENT for the ident plus CHGHOST for the host. valid_vhost validates
the optional ident part; ON/SET/ACTIVATE/identify all route through
apply_vhost, so a bare host still works unchanged. New SetIdent action.
REQUEST <host> records a pending vhost (typed Option<PendingVhost> on the
account, VhostRequested/VhostRequestCleared events); operators review with
WAITING and either ACTIVATE (assign + apply to online sessions) or REJECT
(clear it). Host validation and the operator gate are shared with SET.
New HostServ pseudo-service assigns and applies vhosts. A vhost is a typed
Option<Vhost>{host,setter,ts} on the account (VhostSet/VhostDeleted events),
applied to the displayed host on identify and re-applied by ON; OFF restores
the normal host for the session. Operators SET/DEL (applied at once to
online sessions) and LIST; hosts are validated as dot-separated labels.
New SetHost NetAction -> InspIRCd ENCAP <target-sid> CHGHOST <uid> <host>.
[jsonrpc].tls = { cert, key } makes the endpoint terminate TLS itself,
which also enables HTTP/2 (ALPN), so traffic is encrypted right up to
fedserv — for when it runs on a different host from the site, or the
browser's TLS should reach it directly. Absent = plain HTTP for a
reverse-proxy front. Reuses the same cert/key shape as [grpc]; a crypto
provider is pinned before building the config (the tree carries more than
one). Verified the TLS config builds against a real cert without a
provider panic.
[jsonrpc].origins allowlists the websites whose browser JS may call the
stats API cross-site. The request Origin is echoed back only when it is
on the list (never *), OPTIONS preflight is answered with the allowed
method/headers, and Vary: Origin is always set. Non-listed origins get
no allow-origin header, so browsers block them.
Plain HTTP + JSON-RPC 2.0 (axum, already in the tree) so a site backend
can pull stats without gRPC stubs. Methods: stats.get (counters + gauges)
and stats.channel { channel } (lines + top talkers). Optional [jsonrpc]
config, off unless set.
Security: bearer-token authenticated with a constant-time compare, refuses
to start with an empty token, read-only (only stats.* methods — no way to
mutate state), method-allowlisted, a 64 KiB body limit, and meant to bind
to localhost beside the backend.
Stats now have their own service. StatServ answers /msg StatServ <#channel>
(per-channel activity — lines + top talkers, founder-or-admin) and
/msg StatServ SERVER (the shared cross-service counter registry + gauges,
operators only). The shared counter map moved from the engine onto the
network view so both StatServ (via NetView::stat_counters) and the gRPC
Stats API read one source. BotServ's BOTSTATS command is removed; the
per-channel activity is still recorded in the dispatch loop.
SET <#channel> VOTEKICK <n> enables in-channel voting: members type
!votekick <nick> (or !voteban), each voter counts once, and when n votes
are reached within the window the assigned bot kicks (or bans then kicks)
the target. Votes are ephemeral and lapse after two minutes. Crowd
moderation with no op online.
Responses can now use $1..$9 to substitute regex capture groups (e.g.
ADD hi (\w+)|Hi $1!), and an optional trailing |<secs> sets a per-trigger
cooldown so a trigger can't be spammed. The cache compiles each pattern to
its own Regex (for captures) and tracks last-fired for the cooldown.
BOTSTATS <#channel> reports the lines the bot has seen this session and
the top talkers. Tracked live in the network view (bounded per-nick map),
recorded for bot-assigned channels only, and also fed to the shared stats
registry as botserv.messages.
nickserv.identify / nickserv.identify_fail and chanserv.register /
chanserv.drop are now recorded via ctx.count, so the Stats API reports
each module's own events, not just BotServ's.
A single namespaced counter map on the engine that every service reports
into: ServiceCtx::count(key) is folded in after each command, the engine
bumps its own events (botserv.kick/ban/warn/greet/trigger), and a
per-service command counter (nickserv.command, chanserv.command, …) is
tallied in dispatch. stats_snapshot() merges those with live gauges from
the store (accounts/channels/bots/opers total).
Exposed over gRPC as a new Stats service (Get -> map<string,uint64>),
authorized by the same bearer token as Directory/Accounts — so any
consumer (a dashboard, Django) reads every module's stats through one
pipe. New service is additive to the proto; existing RPCs untouched.
KICK <#channel> DONTKICKVOICES <on|off> exempts voiced users from the
kickers. The network view now tracks +v the same way it tracks +o: a new
ChannelVoice event is parsed from FMODE +v/-v and from +v join prefixes,
and set_voice/is_voiced mirror set_op/is_op (voice is cleared on part and
quit). scan_ops generalised to scan_status for any status letter.
Also scrubbed third-party project name-drops from code comments.
TRIGGER <#channel> ADD <regex>|<response> | DEL <num> | LIST | CLEAR: when
a channel line matches <regex>, the assigned bot says <response> ($nick
becomes the speaker's nick). Only the first matching trigger fires, and a
line the bot just kicked for gets no response.
Same engine machinery as badwords: patterns compile into one RegexSet per
channel (linear-time, ReDoS-safe), cached and rebuilt only on a revision
change. Patterns validated + size-limited at add time.
With WARN on, a user's first kicker offence gets a private warning from
the bot instead of a kick; the next one is kicked (and then TTB can ban).
The warned flag rides the existing per-user chatter state.
Copies a channel's kickers, badword list, greet and nobot settings onto
another channel (founder-or-admin on both). Cheap because the config is a
typed KickerSettings + Vec<String>, not a bag of stringly extensibles.
Reports which content kicker (caps/formatting/badwords) a line would trip,
without kicking anyone — so an op can tune caps% or a badword regex safely.
Reuses the typed KickerSettings::violation() and the shared badword-set
builder (moved to db so the test and the live path share it).
SET <#channel> NOBOT <on|off> reserves (un)assigning a bot for services
operators (the founder is refused). SET <bot> PRIVATE <on|off> (admin
only) marks a bot operators-only to assign; it's flagged in BOT LIST and
the private flag survives BOT CHANGE. Both gates live in the ASSIGN path.
BOT CHANGE <oldnick> <newnick> [user [host [gecos]]] renames or
re-identifies a bot; omitted fields keep the current values. A rename
moves every channel the bot was assigned to onto the new nick. reconcile
now tracks each live bot's identity hash, so a change that keeps the nick
but alters user/host/gecos quits the stale pseudo-client and reintroduces
it (rejoining its channels) — a rename does the same via the registry.
KICK <#channel> TTB <n> makes the bot ban a user after they've been
kicked n times (by any kicker) instead of only kicking; SET <#channel>
BANEXPIRE <duration|off> controls how long that ban lasts. One clear
per-channel knob rather than Anope's per-kicker ttb array.
The kick count rides in the existing ephemeral per-user chatter state, so
it's dropped when the user parts or quits. On the ban the bot sets +b on
an ideal host mask and, if BANEXPIRE is set, queues an unban that is swept
off the next event — no timer thread. Ban time is injectable for tests.
BADWORDS <#channel> ADD|DEL|LIST|CLEAR manages a channel's badword
patterns, each a regular expression, so a channel matches whatever it
wants — an improvement over Anope's fixed ANY/SINGLE/START/END. Enable
with KICK <#channel> BADWORDS ON.
Uses the regex crate: it's finite-automata based with linear-time
matching, so untrusted user patterns can't cause catastrophic
backtracking (ReDoS). A channel's patterns compile into one RegexSet
(single-pass match against all of them), cached in the engine and rebuilt
only when the list's revision changes — so the hot path is one is_match
call, no per-message compilation or allocation. Patterns are validated
(and size-limited) at add time; matching is case-insensitive by default.
KICK <#channel> FLOOD ON [lines [secs]] kicks a user who sends too many
lines too fast (default 6 in 10s); KICK <#channel> REPEAT ON [times]
kicks one who repeats the same line (default 3, case-insensitive).
These are stateful, so the engine keeps per-channel/per-user counters in
an ephemeral map that is never event-logged. Hot-path conscious: a channel
with no kickers costs one lookup and returns; a returning speaker allocates
nothing (get_mut, not entry); repeat comparison is an allocation-free
case-folding 64-bit hash. Counters are dropped when a user parts or quits,
so the map stays bounded. Flood time is injectable for deterministic tests.
The assigned bot now enforces kickers on channel lines: KICK <#channel>
CAPS ON [min [percent]], and BOLDS/COLORS/UNDERLINES/REVERSES/ITALICS
for control-code formatting. DONTKICKOPS exempts channel operators. A
tripping line is kicked by the bot itself.
Kicker config is a typed KickerSettings struct on the channel (event-
sourced like ChanSettings) with a violation() check; the engine runs it
on every channel PRIVMSG. Factored the founder-or-admin gate into a
shared require_channel_admin helper (assign/set/kick).
Members set a personal greet with NickServ SET GREET; a channel founder
enables display with BotServ SET <#channel> GREET ON. On join, if greets
are enabled and the member has channel access and a non-empty greet, the
assigned bot shows [account] greet in the channel.
Per-account greet is a typed field on Account (AccountGreetSet event, rides
the account snapshot); the per-channel toggle is a new ChanSetting::BotGreet
bool. Greet is public in NickServ INFO.
A message beginning with ! in a channel that has a bot assigned is a
fantasy command: !op, !kick, !topic and the rest route through ChanServ's
existing handlers (the word becomes the command, the channel its first
argument), so access checks and behaviour are reused verbatim. The
resulting modes, kicks and replies are re-sourced from the bot, so the
bot is the visible actor.
The channel's assigned bot can now speak: SAY <#channel> <text> and ACT
<#channel> <text> (CTCP ACTION), gated on channel-operator access. Bots
are tracked in the network view so uid_by_nick resolves them, and the
message is sourced from the bot's own uid via a new ServiceCtx::privmsg.
INFO <bot> describes a bot and lists the channels it serves; INFO
<#channel> shows which bot is assigned to a channel. Own file, wired
into the dispatcher.
Match the NickServ/ChanServ layout: lib.rs keeps only the service struct
and the on_command dispatcher, and each command moves to its own file
bridged with #[path]. No behaviour change.
BotServ: bot.rs, assign.rs.
MemoServ: send.rs, list.rs, read.rs, del.rs.
A channel gains a typed assigned_bot (event-logged, snapshotted). ASSIGN
<#channel> <bot> / UNASSIGN <#channel> is founder-gated (or admin oper). The
engine reconcile now also diffs channel assignments against tracked bot
memberships: a live bot IJOINs its assigned channels and PARTs unassigned ones
(new ServiceJoin/ServicePart actions), at burst and after commands; a deleted
bot's memberships are dropped. Engine test for join-on-assign / part-on-unassign.
The engine now reconciles the bot registry against the network: at burst and
after each command it introduces any bot not yet live (minting a SID-prefixed
uid) and quits (QuitUser) any removed from the registry. So BOT ADD makes the
bot appear immediately and BOT DEL removes it; a reconnect re-introduces all
registered bots fresh. Adds set_sid + per-connection bot-uid tracking. Next:
ASSIGN a bot to a channel, then in-channel fantasy commands.
