Coffey Labs is now Coffey Labs LLC, an Arizona limited liability company. Copyright lines and SPDX-FileCopyrightText headers naming Coffey Labs or John Coffey now name Coffey Labs LLC. Upstream copyright notices are unchanged.
Deliverability spec (inbuxa-drafts specs/deliverability.md), the server
side. Every node that sends mail checks itself once a day, at its own
minute in the first hour (UTC), and when an administrator asks:
- its outgoing addresses (the connection strategy's, or what its EHLO
name resolves to), their reverse DNS and whether it resolves back,
and nine blocklists, read by each list's own codes so a refused
query is never taken for a listing (DL-1 to DL-6);
- for every domain: SPF for each address, each DKIM key (by signing a
message that's never sent and verifying it as a receiver would),
DMARC, the MTA-STS policy against the MX, TLS reporting, and the
domain blocklists (DL-7 to DL-12);
- whether it holds a certificate for its EHLO and MX names (DL-13).
It keeps one report per node, facts only; the console grades them.
- inbuxa:DeliverabilityReport: /get, and a create that asks every node
to check now, broadcast as DeliverabilityCheck (DL-15). A tenant
administrator gets their own domains only (DL-20).
- inbuxa:DeliverabilitySettings: which built-in lists are left out, and
the lists themselves (DL-6).
- sysDeliverabilityGet, sysDeliverabilityUpdate, sysDeliverabilityCheck;
a tenant ceiling always turns the last two off.
A school, or any organization that doesn't want people's mailboxes
shared, can now turn that off (multi-account spec, MA-C). Two switches
at two levels, as the legacy-protocols switch has:
- mailSharing: people may share their own mail folders;
- addAccounts: people may add other accounts to the webmail (read by
the webmail's account switcher, MA-B).
inbuxa:SharingPolicy/get and /set hold them: the server's policy has
the singleton id, each tenant's has the tenant's id. Both default to
on, so nothing changes until someone turns one off. A tenant's
administrator changes their own tenant's (the domain's permissions, as
for its protocols switch); only a server administrator with
sysSharingUpdate changes the server's; a tenant can never be looser
than the server (forbidden). Every change goes through the audit log,
and rebuilds every access token, here and on every node.
With mail sharing off for an account's tenant (or the server):
- Mailbox/set and IMAP SETACL refuse to start or widen a share
(forbidden / NO [NOPERM]); narrowing or ending one is always allowed;
- shares already made give nothing while it is off: an access token
leaves out mailbox grants from such an owner. They stay stored, so
turning sharing back on restores them (John, 2026-10-05);
- a lock's and a shared mailbox's grants are an administrator's and
always count, and group membership was never a share.
The session's own account says mailSharing and addAccounts, the
stricter of the two levels, so front ends can hide what is off.
Tests: a new sharing_policy suite with a school tenant, its own
administrator and two people outside it: on by default; the school's
administrator turns it off but can't touch the server's; an old share
stops working and a new one is refused while someone outside the school
is unaffected; a shared mailbox in the school keeps working; the server
off can't be loosened by the tenant; on again restores the old share;
ending a share works while off; and every change is audited. A unit
test covers the stricter-only rule. sharing_policy_tests, jmap_tests,
imap_tests, account_lock_tests and audit_log_tests pass (RocksDB).
A shared mailbox (support@, legal@) belongs to no one person: nobody
signs in to it, and the people assigned open it beside their own mail
at an access level an administrator chose. An account lock already is
most of that: it keeps receiving mail, refuses every sign-in, and its
delegates reach it through real grants on every container (so IMAP,
DAV and JMAP honor them), never including Share. So a shared mailbox is
a lock of a second kind (multi-account spec, MA-S; John, 2026-10-05).
Lock gains kind: "lock" (the default, so stored locks read as before)
or "sharedMailbox", set on create and fixed after. A shared mailbox:
- needs no reason to make, change or end;
- holds up to 100 people, where a lock holds 10;
- runs its own Sieve replies and redirects, so an automatic
acknowledgement goes out (a lock answers no one);
- records only what is sent as it (audit_send_as, which now covers it),
not AL-9's access and per-change records, which would bury the log
for a busy desk;
- sends only as itself (MA-S3): From and Reply-To must be its own
addresses, so answers come back to the mailbox and not to whoever
replied; anything else is forbiddenFrom.
The session marks it delegation: {locked: true, kind: "sharedMailbox"},
so a front end that knows no kind still treats it as a lock. The
console's layout gains Management › Directory › Shared Mailboxes
(CustomComponent/SharedMailboxes).
Tests: the account lock suite now goes on to a shared mailbox: made
without a reason with twelve people, sign-in refused, the session's
kind, its vacation reply delivered, an answer sent as it and recorded
as the agent with no per-change records, and a Reply-To naming the
agent refused; a lock unit test reads a stored lock without a kind.
account_lock_tests, jmap_tests, audit_log_tests and imap_tests pass
(RocksDB).
A group's members reach its mailbox through membership, which counts
as owning the account, so every ACL check was skipped: on a scratch
server a member gave an outsider read access to the group's Inbox with
one Mailbox/set shareWith, with no administrator involved and nothing
audited. Who is in a group is an administrator's decision.
AccessToken::is_group_member_only names that case (in the account
only through a group, without Impersonate). For such a member:
- Mailbox/set with a shareWith change, on create or update, is
refused as forbidden;
- IMAP SETACL and DELETEACL answer NO [NOPERM];
- myRights reports mayShare false, and MYRIGHTS leaves out "a";
every other right stays.
Administrators and the account itself are unchanged. The JMAP ACL
test's group section now checks all three for a member and that the
outsider still has nothing (specs/multi-account.md, MA-D0, G1).
jmap_tests and imap_tests pass (RocksDB). The IMAP refusal has no test
of its own yet; imap_tests passing shows the rest is unchanged.
A group's members can send as the group, and the message says only
From: the group, so nothing recorded which person sent it. Every
submission whose envelope sender belongs to another account now writes
an audit record: the person as actor, an EmailSubmission target named
by the address and owned by that account, and "Sent as <address>",
with ", from <account>" when it went out through the sender's own
account rather than the group's.
A delegate's send is left to AL-9's record, and a send from the
sender's own address writes nothing. No Sender: header is added: the
audit log is where the real sender is named. email_submission_set now
takes the access token, from its one caller.
The audit suite has a group member send once as the group (one
record, with the address, account and details) and once as themselves
(none) (specs/multi-account.md, MA-D0a, G2).
The webmail repository was renamed from ihasmail-inbuxa to inbuxa-webmail
on 2026-10-05. The OAuth client id stays ihasmail-inbuxa: that is what the
server registers, so the backticked and quoted ids are unchanged.
queue.count, user.count and domain.count count the whole cluster, and
only the node with the metrics-calculation role works them out. Every
node still stored them. On the others the queue gauge only moves with
local queue events, so it had drifted below zero (production: node 0 at
18,446,744,073,709,551,596, node 1 at ...613, i.e. -20 and -3), and
the account and domain counts stayed at 0. A reader taking the latest
reading got whichever node wrote last.
sample() now takes whether the node calculates them and leaves them out
otherwise. A unit test covers both cases.
Each node stores histograms as running totals since it started. A sample
didn't say which node wrote it (the node was only in the id's low bits),
so a reader couldn't diff totals per node, and the console diffed across
nodes: on the three-node production cluster the delivery attempt time
read 14.7 s over the last hour against 0.7 s from the nodes' own figures.
x:Metric/get now returns nodeId alongside timestamp, both from the id.
The telemetry suite checks every sample carries it.
Anyone could host a copy of a front end on a server of their own,
collect a person's password there, and replay it as HTTP Basic against
JMAP or the API. Cross-origin rules don't stop that, since a server
isn't a browser, and neither does client registration, since Basic
never goes through OAuth (contract C-23).
JMAP (session, API, upload, download, event source, WebSocket), /api,
/auth/introspect, /auth/userinfo and authenticated /auth/register now
refuse an Authorization: Basic header before looking at the password,
with a 401 whose only challenge is Bearer. A wrong password gets the
same answer as the right one. CalDAV and CardDAV keep Basic, and their
401s still offer it. The sign-in page's /api/auth takes the password in
its body and is unaffected, as is the token endpoint's client
authentication.
Bootstrap and recovery mode accept Basic everywhere, as they keep
permissive CORS. INBUXA_HTTP_BASIC_AUTH=all puts it back everywhere;
dav is the default, and any other value logs a warning and keeps it.
Test builds accept Basic everywhere, since the integration suites sign
in with passwords, and legacy_protocols.py sets the variable.
Tested: unit tests for the paths, and tests/e2e/http_basic_auth.py
against the debug build, 26 checks, including both front ends' sign-in
path and a refused unregistered redirect.
Phase 2 of the journaling spec.
- A copy of each message is taken in MessageWrapper::queue, after DLP and
transport rules, for every enabled journal that takes it (direction and
scope: everyone, or accounts, groups, domains, tenants). If the copy
can't be taken the message isn't queued (temporary failure).
- The journal report: the envelope one field a line (sender, To, Cc, Bcc
from the envelope, list members from their ORCPT, direction, held for
review), then the queued message byte for byte as message/rfc822.
- The built-in journal under J in the inbuxa subspace: one chain per node
whose links name each entry by SHA-256, so entries can expire out of
chain order; purge leaves a marker, and verify catches an entry changed
or removed early and a report that doesn't match.
- Retention per journal (30 to 3650 days); an entry keeps what it was
written with. The daily maintenance purges what's due, keeping entries
whose people a legal hold covers (deleted accounts a hold keeps too),
and records the counts in the audit log.
- inbuxa:Journal get/set, audited by the request layer. Permissions
680-683: administrators see and change journals; the Compliance Officer
sees, searches and exports. Whoever changes journals may grant search and
export without holding them, so officers can still be appointed.
- Catalog entries (inbuxa:Journal, source "journal"); spec as-built notes.
tests/src/system/journal.rs: validation, internal mail with a Bcc,
outgoing into two journals, incoming over LMTP, the report and its
original, tamper and early removal caught, hold-aware purge, retention
changes leave entries alone, disabled and removed journals take nothing.
Phase 2e of the DLP and mail flow rules spec, the API half.
- inbuxa:MailRule/get and /set under urn:inbuxa:jmap. Rules convert
through serde, so what a client sends is the stored format. A create
or change is validated whole (Rule::validate) and refused with the
property at fault; id, createdBy, createdAt and updatedAt are the
server's. Every change goes through the request layer's audit record.
- Six permissions, ids 674-679 (enum and schema labels): mail flow rules
(sysMailRuleGet/Update), DLP rules (sysDlpPolicyGet/Update) and held
mail (sysDlpReviewGet/Update, for phase 3). Either kind's permission
gets through the gate; the handler shows and changes each rule only
with its own kind's. All server-level: a tenant is refused (settled
answer 3).
- Administrators get all six; the server-level Compliance Officer gets
DLP rules to see and held mail to review (settled answer 4), added
once to an existing server's officer role by the grant mechanism,
which gains an officer audience.
- Privacy catalog entry for inbuxa:MailRule.
tests/src/system/mail_rules.rs: create, list in order, validation,
server-set properties refused, update, kind-separated permissions for
an officer, destroy, audit records.
When a valid certificate already covered a domain's names (one stored
by hand before the domain was switched to automatic, for instance), the
renewal task ended with NotDue, which the task manager treats as a
permanent failure. Nothing rescheduled it, so the certificate expired
unrenewed. The renewal now returns a new AcmeRenewal task due when the
certificate falls due, the same way a successful renewal does, and logs
it as a backoff.
The ACME integration suite checks that renewing again right after
issuance hands back one AcmeRenewal for that domain, due at the
certificate's renewal point.
Personal-data catalog spec, default D5 (settled 2026-09-28; built after
the v0.16.24 import's spam-rules loader landed). msbl.org's EBL is sent
a SHA-1 of every email address it's asked about. A new install's first
boot now leaves a note, and the rules update, once the bundled rules
are in, switches STWT_MSBL_EBL_EMAIL off and forgets the note, so it
happens once; the loader keeps that switch through later updates. An
existing server has no note and keeps every blocklist as it is.
Also fixes the data inventory's DNSBL endpoints: a zone is an
expression (`ip_reverse + '.zen.spamhaus.org'`, conditional branches,
`hash(email, 'sha1') + '.ebl.msbl.org'`), and the zone names are now
the quoted literals that start with a dot, from every branch, rather
than the expression's text.
Tested: unit test for the zone rule; the compliance system test (no
note, no change; the inventory lists ebl.msbl.org, not a hash; with the
note the blocklist goes off; the note works once); the system suite;
fork checks.
Personal-data catalog spec, §6 (Phase 3c).
inbuxa:DataInventory/get evaluates the catalog against the server's
live settings and says what this server holds: for each source and each
object that can hold personal data, its categories and whose data it
is, whether it is collected here at all, what bounds its retention (the
live value of the setting that does, or unbounded), whether it leaves
the host and to which endpoints, and a summary. Every host that
receives something is listed once as a candidate processor with what it
receives. Inside a tenant it answers with the tenant's slice and none
of the server's processors. Read-only, with sysComplianceGet.
inbuxa:InventorySnapshot/get is the history: a dated copy of the
evaluated inventory, recorded when it changes -- after a registry write
to an object the inventory reads, after inbuxa's log, audit or AI
settings change, and on the daily clean-up -- and kept as long as the
audit log's records. ids: null lists every snapshot, newest first; the
full inventory only when asked for.
The catalog is embedded and parsed at start (new dependency: toml,
MIT/Apache); the evaluation is a pure function of it and the live
facts, so each configuration is tested without a server. Loopback
endpoints stay on the host; any other configured endpoint leaves it.
Tested: unit tests for the evaluation (a new install's defaults, an
external blob store, a hosted AI endpoint, telemetry off, a tenant's
slice, hosts from URLs, loopback), snapshots, and the fact gathering's
store and duration rules; the compliance system test, extended (the
officer reads the inventory, a plain user is refused, a tenant's
officer sees its slice and no processors, a webhook to another host
becomes a processor and a snapshot names x:WebHook, a retention change
reads through); the system, audit, legal hold and account lock suites;
fork checks. The system suite failed once of three runs with an email
import's blob not found, in antispam.rs; the same happened once in
purge.rs on the previous branch. Nothing here touches uploads; noted
for a separate look.
The schema conflicted as a binary file: taken from main and the one
edit here re-applied (sysComplianceGet after sysLegalHoldExport). The
import kept the permission count at 673, so the new id stays 673.
Retested on the merged tree in its own target directory: the
compliance and system suites pass. One earlier system run failed in
purge.rs (an imported blob not found) and didn't recur.
Personal-data catalog spec, §7 (settled 2026-09-28).
sysComplianceGet (673) sees the data inventory and compliance
overview: superusers and, for their tenant's slice, tenant
administrators, by default and through the one-time grants on servers
that already have their roles stored.
A Compliance Officer role at server level holds it with reading and
exporting the audit log, placing, widening, releasing and exporting
legal holds, seeing account locks, and reading accounts, lists,
domains, tenants and roles. It changes no server setting, creates or
deletes no account, and can't shorten audit retention.
A tenant's accounts can hold only roles of their own tenant (MT-3), so
the tenant role is one "Compliance Officer" role per tenant, without
holds (LH-13): made once for every tenant a server has, and whenever a
tenant is created. While nobody holds it, it is removed with its tenant
so it doesn't block the delete, and put back if the delete is refused
for another reason. Both roles carry a user's own permissions too,
since roles given to a person replace the default user role, which a
tenant's accounts can't hold anyway.
Every server makes these once, new or existing -- the built-in roles
are only made on a server with none -- and records each under P c, so a
role an administrator deletes stays deleted.
Tested: unit tests (neither role changes a setting beyond a user's
own; holds for the server's officer only; per-place records); a new
compliance system test (one server-level role; an officer reads the
audit log, places and releases a hold, and is refused a setting, an
account and audit retention; a tenant gets its role, whose holder reads
the tenant's audit log and no holds; a tenant with an unused role is
deleted and the role goes with it); the system, audit, legal hold,
account lock and SCIM suites; fork checks. The directory suite needs
its LDAP container and wasn't run here.
Personal-data catalog spec, default D1 (settled 2026-09-28): log files
were never deleted. inbuxa:LogSettings.keepForDays says how many days
rotated log files are kept; unset (null) keeps every file, as before,
and a new install sets 30 days.
It is a fork-owned setting, stored under T + l as audit retention is,
not a field on x:TracerLog: that object is also stored inside
x:Bootstrap with a field after it, so a new field would change
x:Bootstrap's stored format. Server-level, with the tracers'
permissions (sysTracerGet, sysTracerUpdate); changes are in the audit
log, before and after.
Log files are local, so every node deletes its own: hourly, and at once
when the setting changes on that node. Only regular files named
<prefix>.<something> in each enabled log tracer's directory, last
changed more than the limit ago, are removed; the file being written is
never that old, and nothing else in the directory is touched. Minimum
one day. The catalog classifies inbuxa:LogSettings and points the log
file's retention at it.
Tested: unit tests for the file rule (only this log's old files; the
current file, other files and directories stay) and a purge on disk;
the system suite, which reads, sets, refuses zero, restores null and
checks the audit records; fork checks.
The schema, which both sides changed, merged as JSON with no conflicts.
The personal-data catalog (#83) gains upstream's new x:DnsServerPowerDns:
nothing personal but its API key, like the other DNS providers.
Personal-data catalog spec, defaults D2, D3, D4, D6 and D7 (settled
2026-09-28, new installs only):
- D2: automatic IP bans expire after 30 days instead of never; D3:
spam training samples, whole messages, are kept 90 days instead of
180; D4: Pyzor, which sends a digest of each message's text to a
public server, is off; D6: delivery history is kept 14 days instead
of 30. Written on the first boot of a new install only -- one with no
roles yet, the same test the built-in roles use -- by reading each
singleton, setting these fields and writing it back whole. A server
with roles keeps its settings, saved or default.
- D7: a webhook created from now on starts with the include policy and
no events, so it sends nothing until events are chosen (Rust default
and schema default, marked). The registry stores every field, so
existing webhooks keep their policy.
- Expired bans are also removed by the daily data clean-up. They
already stopped blocking and were deleted when settings next loaded;
a server that seldom reloads kept them.
D1 (log retention) and D5 (the hashed-address blocklist off) are held,
and the spec says why: x:TracerLog is stored inside x:Bootstrap with a
field after it, so adding one changes that object's stored format; and
the spam-rules loader D5 touches is being reworked by the v0.16.24
import. The spec also corrects finding 3: expired bans were deleted on
settings load; bans were permanent only because no period is set.
Tested: unit tests for the new-install values and that everything else
in each singleton stays; the system suite, whose security test now
purges an expired ban and checks its record is gone; the telemetry
test; common's unit tests; fork checks.
A webhook has a level (info by default) that nothing read: its events
were chosen by its list and policy alone. With the default policy,
exclude, and nothing listed, that meant every event type, including
smtp.raw-input (the raw SMTP bytes, DATA included) and the model's
reply to the spam classifier. The docs suggest a webhook to pass the
audit log to a SIEM; set up that way it would have received whole
messages. Found by the personal-data catalog investigation (finding 1).
Now an include list is sent as named, whatever each event's level:
naming an event is the choice. Otherwise a webhook gets only events at
or above its level, as a tracer does, and never a protocol's raw input
or output (IMAP, SMTP, POP3, ManageSieve, delivery, milter), which
carries whole messages and credentials; those go out only when named.
Tested: unit tests for the rule (level, raw I/O only when named, a
named event below the level, custom event levels, a webhook's own
errors); the telemetry system test, whose webhook names debug-level
connection events and still receives them.
Eight conflicted files resolved, plus the lock file and the schema:
- crates/services/src/task_manager/spam_classifier.rs: upstream's rules
update now replaces existing rules, DNSBL servers, lookups and file
extensions, keeping only whether each is on. Taken, with one difference:
an object an admin edited is kept as it is. Every object an update writes
is fingerprinted (content without `enable`, SHA-256, stored under
SUBSPACE_INBUXA "Sf"), and only one that still matches is replaced.
Scores are never replaced, as upstream has it. The AU-1.10 summary record
now names what was added, replaced and kept, and the bundled rules are
marked applied only when the update fully succeeded, so a failure runs
again on the next start. The marker becomes "3.0.2+2", which runs the
update once on upgrade to fingerprint every rule still as bundled.
- crates/common/src/network/autoconfig/autodiscover.rs: upstream's rewrite
(implicit TLS first, labeled SSL), with the per-protocol switches (LP-7,
LP-14a) passed in as a filter.
- crates/store/src/backend/mysql/{search,write}.rs: upstream's chunked
deletes (no unbounded first DELETE, stop on a short chunk, halve the
chunk on the new chunk-too-large errors) inside the fork's query timeout.
- crates/smtp/src/lib.rs: the fork's queue spawn kept. It already fixed the
stall upstream fixes here (a node without outboundMta stops accepting
mail at about 1024 queued messages), and follows role changes live.
- crates/jmap/src/registry/mapping/bootstrap.rs: the log path stays
/var/log/inbuxa/; upstream's PowerDNS mapping taken.
- crates/main/Cargo.toml: the AGPL-only license kept, version 0.16.24.
- tests/src/jmap/principal/get.rs: the fork's capabilities kept.
- resources/schema/schema.json.gz: merged as JSON; upstream relabeled the
vendor Sieve extensions "(Stalwart)", kept as "(vnd.inbuxa)".
- Cargo.lock: upstream's, with the fork's crates added by Cargo.
Also:
- tests/src/smtp/inbound/spam_rules_kept.rs: an edited rule survives an
update, an unedited one is updated, rules from before fingerprints are
handled, and the audit summary says so. Upstream's own spam_rules test
passes unchanged.
- tests/src/smtp/reporting/reschedule.rs moves to port 19058; upstream's
new spam_rules test took 19057.
- tools/fork/renames.py renames the "(Stalwart)" labels and the default
log path, so neither conflicts again.
- tools/fork/notice-check.py compares against the newest snapshot in the
checked-out history instead of the upstream branch head, so moving the
branch no longer fails other open pull requests.
- tests/src/directory/issuer.rs (since v0.16.23) stays out, and is on the
build check's known list: it tests issuer-based directory routing, which
the fork doesn't have (DIR-2).
- Strip report: docs/fork/strip-reports/v0.16.24.{md,json}.
Upstream commit: af37a234981722493b74623a983581691d2b70b6
Enterprise-only files removed or emptied: 63
Enterprise-only snippets removed: 118 in 50 files
Dangling module declarations removed: 5
Edits turning enterprise off: 25
Third-party code: 14 files, 0 not in THIRD-PARTY.md
Renamed identifiers: 62 in 18 files
Verification: clean
The same Enterprise footprint as v0.16.23. The build check fails only on
tests/src/directory/issuer.rs, unchanged since v0.16.23: it calls a helper
from upstream's Enterprise-only OIDC test, and tests issuer-based directory
routing, an Enterprise feature. main has never carried it.
The legacy-protocols switch was all or nothing. An operator can now stop
POP3 and keep IMAP: each of IMAP, POP3 and ManageSieve has its own
switch, server-wide on inbuxa:ProtocolPolicy and per tenant on
inbuxa:TenantProtocolPolicy (properties imap, pop3, manageSieve).
legacyProtocols stays as the kill-all: setting it sets all three, and it
reads "disabled" exactly when all three are off. A policy stored before
this has only legacyProtocols and reads as all three at that value, so
existing servers and tenants carry over unchanged. In one /set, a
protocol named beside legacyProtocols overrides it.
SMTP submission keeps no switch of its own: sign-in over it is refused
only when all three are off, as the single switch did (LP-6), so
turning one protocol off never stops a mail app sending. For a tenant,
the server's switches and the tenant's count together.
Server-wide, a change closes the listeners of whatever is now off and
puts back the saved listeners of whatever is on again, both in one
change if asked; listeners of a protocol still off stay saved. Sign-in,
autoconfig, autodiscover, PACC (now prepared once per combination) and
the suggested DNS records all follow each protocol separately. A tenant
may turn a protocol on only while the server has it on (LP-9), and the
refusal names which. The JMAP session adds legacyAllowed, the protocols
still allowed for the account; legacyProtocols there keeps its meaning
for older webmail builds. Events name the switches ("pop3 disabled"),
and audit before/after reads every switch even from an older policy.
Tested: unit tests for the switches, the old-policy reading, the
server/tenant combination, the tenant refusal and listener refusal; and
tests/e2e/legacy_protocols.py against a running server, all 100 checks,
including new ones: POP3 alone off closes only its port and refuses
only its sign-in while IMAP and sending go on; only POP3 stops being
advertised; one change closes IMAP and reopens POP3; a tenant turns
POP3 off for itself, and can't turn IMAP on while the server has it off.
inbuxa:LegalHold/get answers accountsCovered, itemsHeld and sizeHeld
when asked: the accounts a hold reaches now (deleted ones it keeps
included) and the archived items it keeps, with their size. Worked out
in one pass over accounts and archive, only for requests that name them.
Held items stay out of the user's quota, as all archived copies do
(LH-9).
Destroying a held account removes the login, as offboarding needs, but
keeps its data as a deleted account with no expiry, whether or not
undelete keeps accounts; its addresses stay reserved and its holds name
it from then on. Destroy-now refuses it, and its DestroyAccount task
defers itself while it's held or its time hasn't come. Holds placed or
released later freeze or free kept accounts in the same settle pass,
with 30 days' grace after the last release (LH-8, LH-10).
Placing or widening a hold freezes what's already archived in its scope
and range, its old deadline noted; releasing one gives each item no other
hold covers that deadline back, or release plus 30 days if later. One
pass over the archive does both and changes nothing twice (LH-6, LH-10,
LH-11). A held archived item can't be destroyed; restoring still can,
and the hold is named only to callers who may see holds (LH-7). Audit
records about a held account survive the purge (AU-7).
Fixes the daily clean-up of expired archived items (UD-13), which never
found any: the registry's unfiltered query reads an all-ids index that
archived items aren't in. Items are now walked account by account, kept
deleted accounts included. Expired items were still removed whenever
their account's archive was read.
Every way of deleting mail (JMAP, IMAP EXPUNGE, POP3, mailbox removal,
Trash emptying) and Sieve scripts, events, contacts and files now asks
how the account's deletions are kept: a hold keeps them with no expiry
(archivedUntil 9999-12-31), even with undelete off; otherwise undelete's
period applies as before (LH-4).
A hold's date range decides by the item's own date (LH-3). Mail is noted
as held at deletion and settled when it's archived, once its received
date is known; outside the range it gets undelete's deadline or isn't
kept. Events go by their start, with a day's slack for time zones;
recurring events, contacts, files and scripts are held whole.
A groupware item's note now stays until its archive succeeds, and a
failure retries the task instead of being logged and lost (LH-5).
A hold reaches an account by name, through any of its addresses'
domains, its groups or its tenant, as they are now, so an account added
to a held domain later is held too. An account that leaves a held
domain, group or tenant stays held: the registry write hook adds it to
the hold by name on every account change, whoever makes it (LH-2).
Server::holds_on answers for the deletion paths, from the store each
time so a hold binds every node at once.
inbuxa:LegalHold get/set places a hold on accounts, groups, domains,
tenants or the whole server, with an optional date range. A hold's range
and scope can only widen, a released hold is read-only, and none is ever
deleted. Placing, changing and releasing each need a reason and are
audited (LH-1, LH-3, LH-10, AU-12).
Permissions 669-672 (see, place, widen or release, export held data)
go to server administrators only; the tenant ceiling always strips them,
as it does Impersonate (LH-13). Schema: Compliance > Legal Holds.
What a hold keeps comes next, through the undelete hooks.
Also moves the lock expiry helpers below the lock module's imports.
A shared account refuses top-level folders, so an organize or full
delegate couldn't add anything to a locked account with no folders. A
delegate who may write now can, as the owner could; the reconcile after
the create grants it the new folder. Read delegates still can't (AL-6,
AL-7).
A delegate's token listed the locked account only for kinds of data it
held grants on, so one with no files (or no calendar) was refused to the
delegate outright: "You do not have access to account". The token now
lists the locked account for mail, calendars, contacts and files alike,
so an empty kind reads as empty. What the delegate may see or change is
still each container's grant (AL-7).
A locked account can't sign in (it fails as a wrong password does), its
sessions end on every node, refresh tokens stop working, and its Sieve
scripts forward and reply to nothing. Mail keeps arriving.
Delegates get real ACL grants on the account's mailboxes, calendars,
address books and files at read, organize or full, with the rights they
replaced restored on unlock. Folders made later are granted after the
create and in a daily sweep. Organize delegates can't destroy; send-as
needs organize or full. The JMAP session marks delegated accounts in
urn:inbuxa:jmap.
New inbuxa:AccountLock object with get/set, permissions 665-668, and a
Compliance > Locked Accounts entry in the schema. Lock, unlock and
delegate changes need a reason and are audited; delegate access and
writes are audited too (audit-hold-lock spec AL-1 to AL-12).
What administrators and the server itself do to the control plane is now
recorded, from inbuxa-drafts/specs/audit-hold-lock.md (AU-1 to AU-12):
settings, accounts, domains, roles and every other registry change, with
each field's before and after (secrets only as "changed"); the fork's own
settings objects; administrator sign-ins (and failed ones to administrator
accounts), master-user and recovery-admin sign-ins, once an hour per
account, method and address; access to another account's data through
impersonation or FetchAnyBlob, once an hour; exports and tamper checks;
and registry writes the server makes on its own, named by subsystem
(system:AcmeRenewal, system:auto-ban, system:directory-sync, ...), with a
spam rules update as one summary record.
No change without its record (AU-3): before a set method changes anything,
a pending record per requested create, update and destroy is written; if
that fails, the method is refused with serverFail. Its outcome follows as
a later entry. A change interrupted by a crash stays "unfinished".
Records live in the fork's subspace under L, as one SHA-256 hash chain per
node. The chain's head is stored, never cached, and every append asserts
it, so two writers can't take the same place. Nothing can edit or delete
a record; the daily purge removes the oldest past the retention (default
730 days, minimum 90) and records where the chain now starts, so
verification still passes. security.audit-recorded (647) copies each
record to webhooks, OpenTelemetry and the log; security.audit-write-failed
(648) reports a failed write.
New JMAP objects under urn:inbuxa:jmap: inbuxa:AuditEvent/get and /query
(filters: time, actor, action, target, account, tenant, outcome, address,
text), inbuxa:AuditSettings, inbuxa:AuditExport (CSV or JSON Lines built
on the server, each line with its chain hash, ending in a manifest; the
created object names the blob and its SHA-256) and
inbuxa:AuditVerification. New permissions sysAuditGet, sysAuditExport and
sysAuditSettingsUpdate: the Administrator role gets all three, the Tenant
Administrator role gets read and export, once, on existing installs too.
A tenant administrator sees records whose actor or target is in its
tenant, including a server administrator's changes there.
Sign-in method on the session: access tokens now remember how they signed
in (password, app password, API key, OAuth client, directory, master user,
recovery admin), including across the HTTP credential cache. New OAuth
access tokens carry their client id in the sealed claims; older ones show
as client "unknown" until they expire.
The schema gains the permissions, the two events and a Management >
Compliance > Audit Log link.
Stack: the request layer boxes every inner future where it's made. Without
that, a debug build overflowed the default 2 MB worker stack on a registry
set; measured with the same request, the branch and main now overflow at
the same stack size (between 1856 and 1920 KiB, debug), so the layer adds
nothing measurable.
Tests: unit tests in inbuxa-features and jmap; system::audit::audit_log_tests
(run with --ignored) passes on RocksDB, SQLite, PostgreSQL, PostgreSQL with a
read replica, MySQL, MySQL with a replica and FoundationDB. The system, JMAP
and SCIM suites pass. authorization.rs skipped fork permissions that guard
no registry object; the audit suite checks a plain user is refused instead.
ai-explain spec, amendment 1 (EX-22 to EX-28):
- answers are three or four sentences, max_tokens 160, cut at 700 chars;
- POST /api/explain streams the answer as server-sent events;
- each node remembers answers in memory (1,000, 24 h), keyed by the facts,
prompt version and model, shared by server-level administrators;
- resources/explain/settings.json.gz ships answers for settings at their
defaults, generated with prepare_setting_explanations (717 for 2026.9.27);
- the system prompt no longer carries the per-request marker, so a model
server can reuse it;
- inbuxa:Explanation gains source, answeredAt and preparedFor.
When a listener couldn't bind its address (a port below 1024 without
root, a port already in use, or the legacy-protocols switch putting a
listener back after privileges were dropped), the bind error was
reported but the socket was still passed to listen(). The kernel then
bound it itself, to a random port on every interface, and the server
logged the listener as started on the port it was configured with.
listen() now refuses a socket that isn't bound, so the listener is
reported with a listen error and skipped, and nothing opens anywhere
unexpected.
A new method, inbuxa:Explanation/set, asks the node's local model for a
short plain-words reading of one thing an administrator is looking at:
a failed recipient in the queue, a Classify verdict, a log line or trace
event, or one setting with its saved value. The server builds the prompt
itself from stored data and the registry schema, never from text the
console sends, and grounds SMTP replies in RFC 3463 and RFC 5321.
What the model is never shown: secrets (including ones nested inside a
setting, like an AI model's HTTP auth), raw protocol events, and the
contents of any other event. A tag name that doesn't have a tag's shape is
refused before a model is asked.
Calls share the AI gate with spam classification, but mail always keeps
its slot, and Explain has its own hourly count per account and its own
on/off switch in inbuxa:AiLimits. The permission is sysAiExplain,
superuser only; tenant administrators can't use it. The session carries
an aiExplain flag so a console knows when to offer the button.
An install whose roles were stored before the permission existed gets it
added once, at start-up, to the roles that are administrators' alone,
not the User role their defaults share with every account. An operator
who removes it later isn't overruled.
Tests: unit tests in inbuxa-features and jmap, and ai_explain_tests
(run with --ignored) covering the acceptance tests and the upgrade.
A cluster rehearsal sent ten x:<Object>/set requests at once and got
ten full reloads on every node. #39's coalescing only joined writes
that queued behind a running reload, but the requests reached the
server about 33 ms apart and a reload takes tens of milliseconds, so
none overlapped one.
A full reload after a registry write now waits for writes to settle:
75 ms after the last one, and at most 250 ms after the first it
covers, so a steady stream still reloads at least four times a
second. 75 ms is a little over twice the gap the rehearsal saw between
requests. A single write pays it once: in the tests a settings write
takes about 140 ms instead of 60. The reload runs in a task of its
own, so a request that goes away doesn't cancel it for the others.
Each write takes the result of the first reload that started after it
was stored (the gate keeps the last 64 results), so applied true or
false still describes the reload that covered that write.
The 33 ms gap was a queue on the server, not password hashing: Basic
credentials are cached per Authorization header, so they are checked
once. Every authenticated HTTP request counted itself against the
account's rate limit by incrementing one counter per account in the
in-memory store, so parallel requests from one account queued on that
key: a row lock on PostgreSQL (a few round trips to the database
each) and conflict retries with a 50-300 ms backoff on RocksDB. An
account with the unlimitedRequests permission (administrators, by
default) passes the rate and concurrency limits anyway, so its
requests are no longer counted. Ten parallel Core/echo calls as the
admin now finish in 1-4 ms; before, they finished one after another
over 20 ms on a local PostgreSQL and 300-450 ms on RocksDB. Other
accounts still count every request.
system::auto_reload::settings_reload_tests: ten concurrent writes now
take one reload (the gate counts them; at most two allowed), all are
applied: true and in the running settings, and a single write takes
exactly one reload. RocksDB and PostgreSQL, 1 reload in 141-196 ms.
With the old behavior (no wait, requests counted) the same writes
took 5 reloads; without the wait but with the rate fix, 2.
cluster::broadcast (3 nodes, PostgreSQL + NATS) and system::reload
still pass.
A cluster rehearsal moved a Log tracer to another directory: the write
was reported x:settingsReload applied:true, but the tracer kept writing
to the old file until a restart. Telemetry::update only refreshed each
running tracer's events, level and lossiness; a tracer's own settings
(path, prefix, rotation, format, endpoint, headers, ...) stayed as built.
Each tracer now carries a hash of the registry object it was built
from, less the fields that change in place. The reload compares it with
the running tracer's: unchanged ones are updated in place as before,
changed ones are started over, new ones started and removed ones
stopped. Only tracers this server started are removed; upstream removed
every subscriber not in the settings, which also cut off live-tracing
streams on each reload.
Starting over is a swap in the collector, so no event is lost or
written twice: a subscriber registered under a running one's id
replaces it between two collection passes. The old one's batch is sent
first (what its full channel can't take moves to the new one), and
dropping it closes its channel, so its task writes what is queued and
ends. Per tracer kind:
- Log: a tracer started over on the same files (rotation or format
changed) waits for the old one to finish, so lines don't interleave.
- Webhook: the task held a sender of its own channel for retries, so
it never ended; retries now use a weak sender, and pending events are
posted when the channel closes.
- OpenTelemetry: pending logs and spans are exported when the channel
closes instead of dropped, and a span that was open across the swap
is exported by the new tracer with the events it saw.
- Console and journal: nothing kept between batches.
- Trace history: built from the tracing store, which takes a restart,
so it is never started over.
No kind needs a restart, so x:settingsReload doesn't gain one.
system::tracer_reload::tracer_reload_tests (new): a Log tracer created
over JMAP writes to its directory; its path is changed over JMAP while
2000 numbered events are emitted; after the reload, events land in the
new file and not the old one, each numbered event is in exactly one of
the two files, and a destroyed tracer writes nothing. On main the new
file never appears.
In cluster rehearsal 3, turning outboundMta off on node1's role was
reported applied (x:settingsReload applied: true), yet node1 kept
delivering mail, a report message included, until it was restarted.
The queue and report managers were started at boot only when the
node's role included outboundMta (crates/smtp/src/lib.rs), and the task
manager only when the role had some task type (spawn_task_manager).
After that nothing looked at the role again: a queue manager that was
running kept claiming and delivering, and one that wasn't never
started.
They now start on every node (outside recovery mode) and follow the
role live:
- Queue manager: before each scan it reads the role from the running
settings. Without outboundMta it claims nothing new; deliveries
already running finish and report back as usual, which releases
their locks. When the role comes back (a reload wakes the manager
with ReloadSettings, and it looks again every 30 s regardless) it
logs queue.started and scans the whole queue at once.
- Report scheduler: DMARC and TLS report events are handled only while
the role has outboundMta, as at boot; events arriving without it are
dropped, as they were on a node started without the role.
- Task manager: task_enabled already read the current role on every
scan. It now also runs on nodes whose role has no task type (the
scan returns at once until one is added), a job claimed before a
role change is handed back at once rather than run or held until
its lease lapses, and a settings reload wakes the manager so a role
that gained task types starts claiming them straight away.
Starting the queue manager on every node also drains the queue channel
on nodes without outboundMta. Upstream left that channel unread, so
each message queued there parked a refresh in it, and by the code,
queueing would block once 1024 had piled up (not reproduced here).
A role object edit reaches the nodes that name that role in
INBUXA_ROLE. Moving a node to another role still means changing its
environment, and so a restart. Listener changes in a role still need a
restart too (listeners bind at boot); this change is about tasks and
delivery.
cluster::live_roles::live_role_tests (new; PostgreSQL, two nodes over
one store):
1. A node started with outboundMta delivers and runs a TLS report
task; after its role loses outboundMta and the settings reload, a
new message isn't attempted and a new report task stays pending;
with the role back, both are taken up.
2. A node started with no task type at all gains outboundMta: a
waiting message is attempted and a report task runs.
On main the test fails at step 1 ("delivery attempted without
outboundMta"); with step 1 bypassed, step 2 fails (nothing picked the
message up in 20 s).
Cluster rehearsal 3: with PostgreSQL paused (docker pause, so its
kernel still answered TCP keepalives), requests on connections already
checked out hung until it came back, and /healthz/ready stayed 200
through the outage. #41 bounded getting a connection, not using one.
Client-side query limits (store::backend::query_timeout). Every
operation on a PostgreSQL or MySQL connection now runs under a time
limit. A server-side statement_timeout (or MySQL's MAX_EXECUTION_TIME,
which covers SELECTs only) can't do this: the server that would enforce
it is the one not answering. When an operation runs out, its connection
is closed instead of pooled, since a query may still be in flight on it
or a transaction open: deadpool's Object::take on PostgreSQL;
Conn::disconnect on MySQL, which marks the connection closed before it
sends anything, so the pool discards it even when the server never
answers.
- query, 2 minutes: reads, writes (the whole transaction with its
retries), blobs, SQL lookups, search queries and indexing. These take
milliseconds; two minutes leaves room for a large blob over a slow
link and still ends a hang.
- maintenance, 30 minutes: range deletes (account removal, purges),
unindexing, purge_store, and creating tables and indexes at startup,
which can legitimately run long in one statement. Their existing
chunked fallback for server-side statement timeouts is unchanged.
- iterate (exports, reindexing, maintenance scans) can run for hours,
so the query limit bounds each wait for the database (preparing, the
query starting, the next row) rather than the whole scan.
The limits are fixed, like the pool timeouts; the DataStore schema has
no field for them. Tests set them with Store::with_query_timeouts
(test_mode only).
Readiness. /healthz/ready answered 200 whenever a data store was
configured. It now reads one key from the data store with a 2 s limit
and reuses the answer for 2 s, so probes can't load the database;
while one probe runs, others get the last answer. The first failed
probe of an outage is logged. /healthz/live stays 200: restarting a
node doesn't bring its database back, and an orchestrator restarting on
failed liveness would restart every node at once. The container
HEALTHCHECK already uses /healthz/live.
Tests, store::pool_timeout (a proxy that stops forwarding while
keeping connections open plays the paused database):
- postgres_query_timeout, mysql_query_timeout (new): with four pooled
connections open, a read, a scan and a write each fail with "Query
timed out" 2.0 s after the pause (2 s test limit); once the proxy
forwards again the store answers. With the limits set to an hour
(upstream's behavior), the read was still waiting at the test's 20 s
limit.
- postgres_readiness (new, STORE=PostgreSql): a node's data store
goes through the proxy; /healthz/ready is 200, 503 about 4 s after
the pause while /healthz/live stays 200, and 200 again about 2 s
after it ends.
- postgres_pool_timeout, mysql_pool_timeout: pass as before.
store::store_tests (PostgreSql, MySql, including the MariaDB statement
timeout step) and store::task_locks (PostgreSql) pass;
store::search_tests (PostgreSql) fails at the same ordering assertion
(query.rs:684) as on main.
write_reload_target sent AllowedIp writes to the blocked-IP reload, but
that reload rebuilds only BlockedIps. Allowed IPs are parsed into the
core's security settings (Security::parse), which only a full reload
rebuilds, so an AllowedIp write reported x:settingsReload applied: true
while the change wasn't live until the next full reload.
AllowedIp now maps to the full reload, like the other settings objects;
BlockedIp keeps its targeted reload.
system::auto_reload::settings_reload_tests now creates an allowed IP
over JMAP and checks that is_ip_allowed sees it with no ReloadSettings,
and that destroying it takes it out again. On main it fails ("allowed
IP not in the running settings").
A 3-node rehearsal (PostgreSQL + NATS + Garage) found two ways a crash
leaves work stuck:
Pool hangs. The PostgreSQL pool (deadpool) was built with no timeouts,
so a request waited for a free connection, and for one to be opened or
recycled, for as long as it took: forever when the server stopped
answering. MySQL's pool (mysql_async) has no wait timeout at all.
- PostgreSQL: wait 30 s (or the store's timeout if longer), create the
store's timeout or 15 s (it bounds the whole handshake, where
tokio-postgres's connect_timeout covers only the TCP connect), recycle
10 s. The pool config is now always set, not only with
poolMaxConnections.
- MySQL: every connection is taken through MysqlStore::conn(), which
gives up after 30 s.
- Both: TCP keepalive after 60 s idle, so a server that vanished
without closing the connection is noticed in minutes rather than the
two-hour system default.
The DataStore schema has no pool timeout settings, so these are fixed
defaults; the store's own timeout bounds connecting on PostgreSQL.
Task locks. A task lock lasted an hour, so after a hard crash the dead
node's tasks waited up to an hour and five minutes. The lock is now a
five-minute lease: while this node runs a task, the task manager renews
its lock every third of the lifetime (InMemoryStore::renew_lock, a
compare-and-set on the store backends and SET XX EX on Redis, which
leaves a lock that already expired alone). A killed node's tasks run
elsewhere within about five minutes plus the claim recheck. A task this
node holds isn't handed to a worker again by the scan.
store::pool_timeout (new): a local listener that accepts connections
and never answers plays a hung server; a PostgreSQL store with a 2 s
timeout returns an error in about 4 s, and a MySQL store in 30 s.
Without the timeouts both wait for good. store::task_locks gains a
task held for 1.5 lock lifetimes: its lease is still held, and released
when the task ends.
A 3-node rehearsal found that saving an MtaDeliverySchedule left it
unknown to the queue ("Queue strategy not found") until someone ran
x:Action ReloadSettings; only Directory and Authentication writes
reloaded (DIR-17). The admin UI has to remember a separate reload after
every save, and a script or API client that doesn't gets a server
running stale settings.
x:<Object>/set now reloads the running settings when it created,
updated or destroyed an object they are built from, and broadcasts the
same RegistryChange::Reload over the coordinator as ReloadSettings, so
every node applies it:
- Settings objects (MTA, spam filter, listeners, tracers, Sieve system
scripts, cluster roles, directories, ...: the object types the core,
telemetry, listener and directory builders read) get a full reload.
- Certificates, lookup stores and blocked/allowed IPs get their own
targeted reloads.
- Accounts, domains, roles and other data read as needed, stores (they
take a restart) and applications (their own reload action) get none.
Full reloads are coalesced: a write waits for a reload that started
after it was stored and joins one if it can, so a burst of writes, or
a request with many objects, costs one or two reloads, not one each.
The write itself is never undone. When the reload is refused (build
errors in objects that were working, the rule from the previous
commit), the set response says so in a new x:settingsReload field,
{"applied": false, "description": "Saved, but the running settings
were not reloaded. <object>: <error>"}; {"applied": true} otherwise.
The field is absent when the write needs no reload. The description
helper is shared with ReloadSettings' refusal.
Each reload sends the queue a ReloadSettings event, so the SMTP test
harness's read_event, try_read_event and assert_no_events now pass over
those; expect_reload_settings still waits for one.
system::auto_reload::settings_reload_tests (new): an MtaVirtualQueue
and an MtaDeliverySchedule created over JMAP are in the running
settings with no ReloadSettings, and gone once destroyed; eight
concurrent creates all land; a write whose reload fails is stored and
reported applied: false with the error; a domain write carries no
x:settingsReload. On main the new schedule is missing. The cluster
broadcast test (three nodes, PostgreSQL + NATS) now checks that every
node has a schedule created on node 0 without a reload.
jcoffey-dev