Security fix, outside the inbound antispam feature. Found by its
technical review (Opus, 2026-10-12), reproduced on the test stand
before the fix: over an application's level-2 limit, two MAIL FROM on
one connection were answered 451 and then accepted.
Cause. Postfix introduces a client to a milter once per connection.
The milter library (go-milter 0.4.1, session.go) replaces the milter's
session after any reply that is not "continue" - a refusal, or the
accept that used to end a message. The new session never saw Connect,
so it had no client address, and both checks passed a message without
one: overLimit returned early ("no client IP to key on") and
authIPAllowed did the same. So every message after the first on a
connection, and every MAIL FROM after a refused one, skipped the
level-2 limit and an application's client IP list. In the code since
1.9.0 for the list and longer for the limit.
Fix.
- build/postfix-config.sh: milter_mail_macros carries {client_addr},
so the address comes with every MAIL FROM; MailFrom reads it (both
spellings of IPv6) in the journal milter and the inbound journal
milter.
- overLimit no longer asks for a client address: a ceiling is on a
domain or an application.
- authIPAllowed refuses an application that has a client IP list when
no address is known.
- Body replies "continue" in both milters, which lets the message
through as "accept" did and keeps the session.
The inbound journal milter had the same cause with a smaller effect:
rows of a connection's later messages had no client address.
In the same file, the checks of INBOUND_ANTISPAM_MILTER and
INBOUND_JOURNAL_MILTER_SOCKET refuse any character outside their set
first: a glob's "*" matched anything after the first characters.
Tests: a session that never saw Connect is judged by the macro, a
listed and an unlisted address, no address, not an address; the limit
with no client address; the inbound milter answers only "continue".
test/e2e sends two MAIL FROM on one connection over the limit and
wants both refused.
Checked on the test server: e2e red before the fix on exactly that
check, green after; go vet ./... ok, go test ./... ok (22 packages).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
From the security review of 2026-10-12: the allow list matches the
envelope sender, which nobody authenticates. Help's Filter lists topic
says so and points to client addresses and networks.
check.py: 32 screens, 0 violations.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The owner accepted Inbound > Filter lists on 2026-10-12 ("Да"), after
a deny entry he added refused my test message on the live test server
with the list's own reply (554 5.7.1 Refused by this server's deny
list), mail.log carried the milter-reject line the count reads and the
journal its row.
Steps 8-10 of the plan were done with each earlier step and are ticked
with where. 10/12: the technical and the security review remain.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Follows the design commit before it (owner's words quoted there).
The lists' tables are Value, Kind, Matched and Remove. The kind tag
opens the entry's sheet: note, day added, how often it decided a
message in the days the log keeps, and the last match.
Matches are read off the journal on each view, nothing is stored
(inbound.Lists.Usage, store.ListSpamListHits). A row counts for an
entry when the list decided the message - a rejected row whose reply is
the deny rule's message, an accepted row whose verdict carries the
allow rule's symbol for that kind - and the entry fits the row's sender
or client address the way rspamd matches it: an address whole, a domain
as the sender's domain, a network containing the client (an
IPv4-mapped address as IPv4). A failure to count shows the lists with
"never" and logs why.
For that the journal milter keeps the lists' own symbols ahead of its
six-symbol cap: rspamd writes symbols heaviest first, so an allow
entry's -100 came last and was cut behind any six others.
The plan gains "Filter list usage": what counting from the journal
implies, and - the owner asked for it in writing - the conditions under
which a counter kept with each entry becomes worth a migration (the
journal cut by its row cap and not by age; a short retention; counts
that must outlive the journal; a slow page). multimap.conf says which
of its texts SelfPost reads back.
Tests: Usage over every kind on both lists and the rows that are
nobody's; the milter's symbol order; the store query; the page and its
sheets; a handler test from journal rows to the numbers on the page.
Seen: the fixture page with a sheet open, headless Edge at 1360 px -
the same as the accepted drawing.
Checked on the test server, uncommitted tree: go vet ./... ok,
go test ./... ok (22 packages), make e2e ok.
Not seen in a real run: a count above zero. It needs an entry added in
the panel, which I cannot sign in to, and then mail that the list
decides; in particular the deny rule's message arriving in mail.log as
the reply of a milter-reject line is what the trial with rspamc showed,
not yet what Postfix logged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Owner, 2026-10-12, on Inbound > Filter lists as built: "Note хочется
вынести в лист, подсказку, раскрывающийся блок. Тип в отдельную колонку.
Ещё полезно будет дать статистику использования правила, чтобы удалять
устаревшие." Drawn, then accepted ("Да") together with counting the
matches from the Inbound log.
- Kind is a column of its own; the tag no longer wraps under the
value.
- Note and Added leave the table for the entry's sheet, opened by the
kind tag (of his three - sheet, tooltip, disclosure - the sheet: it
exists and works under a finger).
- Matched: how often the entry decided a message, or "never". The
sheet adds the window ("in the last 90 days") and the last match.
A fifth column for the last match did not fit half the page width.
Help, Filter lists: what Matched counts and what "never" suggests.
check.py: 32 screens, 0 violations; outlines.json unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Step 7/12 of docs/plans/inbound-antispam-panel.md. A full backup walks
all of /data, so data/antispam/ was in it from the day the directory
existed; TestCreateIncludesStateExcludesTransient now holds that. On
the first start after a restore the maps are written again from the
database (mailStack.Resync, already there). The guide says what is not
in a backup: the antispam/ directory with the compose fragment and the
rspamd configuration, which is installation and not state.
The plan records the owner's acceptance of Inbound > Log (2026-10-12,
"Этот экран утверждаем") and what he changed on the way to it.
go test ./internal/backup/ ok.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Follows the two design commits before it (owner's decisions quoted
there).
Kit: a new component, Sheet / SheetRow (components.go), partials
sheet_tag and sheet (components.html), its rules in panel.css as in
theme.css, and its sample on the kit page next to status_tag. It is
the browser's popover: no script.
Inbound log: the Why column is gone; the decision tag opens a sheet
with who decided, then either the filter's verdict and its symbols one
a line (a detail the journal milter wrote) or the reply (a detail the
tailer read from mail.log - it starts with its code), and the subject
whole. The sheet's id is the journal row's. A row Postfix itself
refused no longer links to the filter lists.
Tests: the page fixture and view tests for the sheet's markup and
content (TestWhyRows), the link rule, three Deny and two Allow on the
mockup's six rows; e2e reads "Decided by" and "Reply" from the sheet
and wants no sender link on the refused row.
Seen: the fixture page with a sheet open, headless Edge at 1360 px -
the same as the accepted drawing.
Checked on the test server, uncommitted tree: go vet ./... ok,
go test ./... ok (22 packages), make e2e ok.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The sample sheet on the components page still held facts; it shows the
table of name and value the accepted sheet is. In in-log a refused
message's sheet has no Subject row - there is no subject to show whole.
check.py: 32 screens, 0 violations.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Owner, 2026-10-11/12, on the Inbound log as built:
- "Why сделать не столбцом, а листом поверх." Opened by the decision
tag ("Тег."), drawn first, then: its content as a table ("В лист
может быть таблицей?" - "Да. Но хорошо бы ещё её отцентрировать."),
with a narrow column of names and only what the row does not show.
- "Почему Deny у того, кто rejected?" - "Убирай": a message Postfix
itself refused offers no Deny sender. Accepted mail offers Deny,
mail the filter refused offers Allow.
A new component, sheet (theme.css, shown on the components page next to
status_tag): the browser's own popover, so no script - the tag is a
button with popovertarget; Esc, a click outside and the cross close it.
It rises from the bottom edge across the window; its content is a table
of name and value in a 44rem column in the middle. A tag that opens a
sheet carries a dotted underline to tell it from a plain tag.
in-log: the Why column is gone and rows are one height. A sheet holds
who decided, the verdict and the symbols (or the reply), and the
subject whole.
Guard: check.py WIDTH_OK gains ".sp-sheet" - the one component laid out
by the window and not by the page. The owner accepted the drawing with
that said.
check.py: 32 screens, 52 sp- classes, 0 violations; outlines.json
unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Owner, 2026-10-11, of "From · client": "Client в названии столбца
зачем? Мне кажется, оно лишним." The client address stays under the
sender, in mono and muted; an address of that shape needs no label.
check.py: 32 screens, 0 violations; outlines.json unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Step 6/12 of docs/plans/inbound-antispam-panel.md. Written by a Sonnet
subagent from the mockups, reviewed and corrected by me.
GET /inbound/log - the journal a page at a time (50), filtered by
domain, decision and day; QueryInboundSpam / CountInboundSpam with the
viewer's inbound domains as the scope. The global role reads all of it
and gets "Deny sender" / "Allow sender" links into the lists; a user
with the domain role reads the rows of the inbound domains assigned to
them, without the links. The plan said global-only for the journal;
the accepted mockup and 2.0's delegation of Inbound say otherwise -
recorded in the plan as built.
GET, POST /inbound/filter-lists and POST .../{id}/delete - global role,
404 otherwise. Add form (prefilled from the query), the two lists with
Remove as the kit's row action, and "Where the lists go" with the
filter's state and when the maps were last written
(Lists.LastWritten). A refused or duplicate entry is 422 with the
message and what was typed.
Server > Settings: "Keep inbound log rows, days" with the relay on.
Help: Inbound log, Filter lists. Navigation: Log for everyone who sees
Inbound, Filter lists for the global role (Shell.ShowFilterLists).
Status tags accept / reject / tempfail join the kit's mapping.
Tests: page fixtures for both screens, held to outlines.json; view and
handler tests (both roles, filters, paging, relay off, add / duplicate
/ invalid / delete, map file content, settings field, help). e2e reads
the journal from the page instead of the database and adds and removes
a deny entry through the panel, reading deny_domain.map after each.
Seen: both pages and Settings rendered from their fixtures and compared
with the mockups side by side (headless Edge, 1360 px) - same layout
and wording.
Checked on the test server, uncommitted tree: go vet ./... ok,
go test ./... ok (22 packages), make e2e ok.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Guards and mockups only, ahead of the pages (inbound-antispam-panel
6/12). The owner agreed the feature ("Следующая по дорожной карте -
inbound-antispam-panel" / "Да", 2026-10-11) and started this step
("Начинай"), so the two screens are no longer "drawn but not built":
- guard_panel_test.go: in-log and in-filter-lists removed from
deferredScreens. From here TestLegacyRatchetMatchesTheEngine wants
the engine to have both pages - red until the next commit.
- guard_routes_test.go: the four routes join routes.inbound, so they
must exist with the relay on and answer 404 with it off.
Two mockup details brought to what the kit can draw, found when the
pages were built from it:
- help: the two new topics are one paragraph with a break in it, as
every topic is (panel.css has no gap between paragraphs of a topic);
- in-filter-lists: the directory in "Files" is mono at the text size -
the kit's inline code has no small variant.
check.py: 32 screens, 0 violations; outlines.json unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Mockup-first step of inbound-antispam-panel 6/12. The two screens were
drawn for the redesign before the feature existed; this brings them to
what steps 3-5 built.
in-log: the lead says where rows come from and how long they are kept;
a help link in the box head; Why is the engine and the detail as the
journal holds them (rspamd: action, score / threshold, weighted
symbols; postfix and a refusing filter: the reply text). No row
invents what the journal cannot hold: the "selfpost - upstream timed
out" tempfail and the Review link to a quarantine page that does not
exist are gone, a refused message has no subject, an allow-listed one
shows SELFPOST_ALLOW_FROM(-100.00).
in-filter-lists: Remove is a POST - the kit's row action, not a link;
a domain entry is shown as stored, without "@"; "Where the lists go"
names the directory and its six files instead of allow.map / deny.map,
says rspamd rereads them by itself and that Deny wins.
settings: "Keep inbound log rows, days" in the Log retention box.
help: two topics under Inbound, Inbound log and Filter lists, and a
sentence in Settings.
outlines.json rewritten: Help's Inbound box has six topics, was four.
check.py: 32 screens, 0 violations.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Step 5/12 of docs/plans/inbound-antispam-panel.md.
Store: AddSpamListEntry, ListSpamListEntries, DeleteSpamListEntry.
internal/inbound/lists.go validates and normalises an entry - sender
address, sender domain, client address or network (canonical text, an
IPv4-mapped address as IPv4, /0 refused) - and writes the lists as six
map files under /data/antispam (ANTISPAM_DIR), each atomically, 0644 in
a 0755 directory: the filter is another user in another container. All
six are written on every change and on every start with the relay on,
empty ones included. An entry whose maps could not be written is taken
back out of the table. The writer refuses a stored value with
whitespace, a newline or # in it.
The plan's allow.map / deny.map became one file per list and entry
type: to rspamd an IP map and a sender map are different map types.
deploy/antispam/rspamd/multimap.conf holds the rules; the fragment
mounts it and ./data/antispam into the sidecar. Deny is a prefilter
reject, allow a symbol weighing -100. Tried in a throwaway rspamd 3.11
container on the test server:
- with allow as a prefilter accept, an address on both lists was
accepted and an IP on both was refused - the rule that ran first
decided; with the weight, deny won in all five mixed cases;
- IPv4, IPv6, address (any case), domain (not its subdomains) match;
- a rewritten file is reread in about ten seconds, no restart;
- files that appear after rspamd started are loaded the same way.
No page edits the lists yet (step 6).
Checked on the test server, uncommitted tree: go vet ./... ok,
go test ./... ok (22 packages), make e2e ok - with the relay on
/data/antispam is 755 and holds the six maps, 644, empty.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Step 4/12 of docs/plans/inbound-antispam-panel.md.
The log tailer writes an inbound_spam_log row for mail port 25 refused
(internal/logtail/inbound.go): a "reject:" line of the port-25 smtpd is
Postfix's own policy; a milter-reject / milter-hold / milter-discard
line of cleanup is the filter's verdict. Only a recipient at an inbound
domain is a row, so "Relay access denied" for a stranger's domain is
not. The row is dated by its log line; the reply text is cut to 255
bytes of printable ASCII; a reply starting with 4 is a tempfail.
To tell port 25 from 465 / 587 the listener now logs as
postfix/inbound (syslog_name). cleanup logs under one name for every
listener, so its verdict counts only for a queue id the port-25 smtpd
was seen to open (4096 remembered). Lines are matched from the head,
after timestamp and host: their tail is the sender's own text.
Retention: PruneInboundSpam every six hours - older than the
inbound_spam_log_retention_days setting (seeded from
INBOUND_SPAM_LOG_RETENTION_DAYS, 90, range 7-365), then down to 10 000
rows. The setting has no field in the panel yet.
Known limits, in architecture.md: a log line names one recipient, so a
message refused at end-of-message is one row; a line re-read after an
unclean stop would be a second row.
Checked on the test server, uncommitted tree: go vet ./... ok,
go test ./... ok (22 packages), make e2e ok - the unknown recipient at
the inbound domain refused on port 25 has its row (reject by postfix),
the stranger's domain has none.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A candidate, recorded at the owner's word (2026-10-11, screenshot of
Server > Health on selfpost.mixfed.ru): the body is laid out in columns
while the navigation is the phone stack with every section open. The
item states what was seen and a first guess at the cause (980 px falls
between Bulma's 769 and 1024 px breakpoints); nothing is investigated
yet.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Step 3/12 of docs/plans/inbound-antispam-panel.md.
The panel serves a second milter while the inbound relay is on
(internal/milter/inbound.go, socket INBOUND_JOURNAL_MILTER_SOCKET,
default /run/selfpost/inbound-journal.sock). build/postfix-config.sh
puts it last on port 25, after the optional antispam milter, with
default_action=accept; postfix-wrapper.sh waits for its socket. At
end-of-message it writes one inbound_spam_log row per recipient at an
inbound domain - other port-25 recipients (DMARC report addresses) get
none - and always accepts. With a filter attached it reads the verdict
from X-Rspamd-Action and X-Spamd-Result (action, score, up to six
weighted symbols, only what the patterns recognise, 255 bytes); without
one those headers are the sender's and are not read. The plan named
X-Rspamd-Score, which rspamd does not have.
The antispam compose fragment mounts rspamd/milter_headers.conf into
the sidecar so that it adds the two headers (checked in rspamd 3.11:
both routines exist and remove a sender's own headers of those names).
Store: InsertInboundSpam, IsInboundDomain.
Checked on the test server, uncommitted tree: go vet ./... ok,
go test ./... ok (22 packages), make e2e ok - the relayed message of
inbound_relay_delivers_and_refuses has its journal row (accept by
selfpost) and the DMARC report sent through the same listener has none.
Not yet seen in a real run: a verdict written by rspamd reaching this
milter. The e2e stand has no filter, and the test server's only inbound
domain has no upstream, so it accepts no mail.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Step 2/12 of docs/plans/inbound-antispam-panel.md, the first migration
after the 2.0 baseline (2.0.1 was cut before it, as agreed).
0002_inbound_spam_log.sql adds inbound_spam_log (one row per recipient:
domain, client IP, envelope, subject, decision, engine, detail) with
indexes on created_at and (inbound_domain, created_at), and
inbound_spam_list (allow / deny entries, unique per list). DDL only;
decision, engine, list_type and entry_type carry CHECKs, as the
baseline does for its own enumerations. Nothing reads or writes the
tables yet.
Tests: TestBaselineSchema pins both tables and user_version 2;
TestOpenMigratesA20Database opens a file holding the baseline alone and
finds its user kept; TestInboundSpamConstraints.
go test ./internal/store/ ok; go vet ./... ok.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A patch on 2.0.0, all of it found on the test server after the cut:
- a message that is not a report this relay keeps is dropped at once
instead of being retried for days;
- a report is bounded before it is stored, and reports are kept per
domain (security review of the ingest path, nothing Critical or High);
- Full backup from the panel works (it had answered 200 with ten bytes
since 1.3.0);
- no backup holds the Postfix queue, and the panel says why.
Cut: CHANGELOG [Unreleased] -> [2.0.1]; image pin 2.0.1 in compose,
README, guide; version table.
On the test server, on dd0fefd: go vet ./... ok, go test ./... ok
(22 packages), make e2e ok.
Full detail: CHANGELOG.md (2.0.1).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Owner, 2026-10-11: the private key is needed, and a backup from the
panel and one from the console have to be the same.
The commit that made the panel's backup work skipped a file under
certs/ the panel could not open - on the test server the private key,
which the certificate hook leaves to root and Postfix's group. That gave
an archive with a certificate and no key, and two kinds of archive.
- Params.Unprivileged is gone. Nothing is left out on account of who
makes the archive; the queue is out for everyone (previous commit).
- When the panel cannot read a file under certs/, backup.Check says so
before the download and the panel refuses with what to do on the host:
chgrp 998 certs/privkey.pem && chmod 640. backup.UnreadableCert tells
that case from any other failure.
- 998 is the container's selfpost group, which holds the panel and
Postfix. Its id was whatever groupadd gave; the Dockerfile now fixes
it at the value it has always had, because an operator addresses it
by number from the host.
- Guide: method 1 is the same archive as method 2, with the permission
the key needs and a note for whatever renews the certificate. Help and
the box follow the design commit before this one.
- e2e: the restore scenario first makes the key unreadable and expects
the refusal naming the file and the way out, then takes the backup and
expects the key in it; nothing is put back by hand before the restart.
go vet ./internal/backup/ ./internal/web/handlers/: ok
go test ./internal/backup/ ./cmd/... ./internal/web/...: ok
make e2e on the test server: ok
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Owner, 2026-10-11: the key is needed and the panel's archive must be the console's. The box no longer says a certificate file may be left out; Help says the key has to be readable by the panel (group 998 on the host) and that the panel refuses, naming the file, until it is.
check.py: 0 violations; outlines.json unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Follows the design commit before it: the paragraph of the Full backup box and the Backup topic of Help.
go test -count=1 ./internal/web/...: ok
Backup page at 1280: no overflow, the box holds the longer paragraph.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Owner, 2026-10-11: the user has to be told why. The box gives the reason in a sentence - the queue cannot be copied reliably while the server is running; the Help topic explains it (Postfix keeps moving messages, a copy could hold one twice or miss it), what follows (waiting mail is not carried over, a restored server starts with an empty queue) and what to do to keep the queue (stop the container, archive the project directory).
check.py: 0 violations; outlines.json unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Owner, 2026-10-11: the queue goes out of the command's archive too.
The commit that made the panel's backup work left the queue in the
archive of selfpost-backup, which runs as root and can read it. But that
copy was never the queue: the archive is made while the instance runs,
Postfix moves messages between the queue's directories all the time, and
a message on its way could be in the archive twice or not at all. One
rule now - QueueDir is left out of every archive - and
Params.Unprivileged is only about a certs/ file the panel may not open.
What is not carried over is mail still waiting for delivery; a restored
instance builds an empty queue (the e2e restore scenario starts one).
To keep the queue: stop the instance and tar the project directory, as
the guide's methods 3 and 4 already describe.
Guide: the description of a full backup, methods 1 and 2 and the
comparison table. Help follows the design commit before this one.
go vet ./internal/backup/: ok
go test ./internal/backup/ ./cmd/... ./internal/web/...: ok
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Owner, 2026-10-11: the queue is left out of the command's archive as well as the panel's. The Help topic no longer says selfpost-backup takes it; the command still takes the private key the panel cannot read.
check.py: 0 violations; outlines.json unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
After the rate-limit checks the run takes a full backup through the
panel, reads the archive through while the container can still say what
went wrong, then removes data/, certs/, docker-compose.yml and .env,
unpacks the archive into the project directory, puts the private key
back (the panel's archive comes without it and without the queue, both
asserted), and starts the stand again.
What has to hold: manifest.json is consumed by the first start, the
administrator signs in, the domain shows the same DKIM value, and a
message sent with the application's old password reaches the sink and
verifies against the record published before the backup.
The port-25 scenarios then run on the restored installation.
make e2e on the test server: ok.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Found by the e2e restore scenario of the next commit: the Full backup
button answered 200 with ten bytes. The archive includes the Postfix
queue, Postfix keeps every queue file and directory readable by its own
user alone (not configurable), and the panel runs as its own user - the
walk stopped at /data/postfix/queue/active with "permission denied"
after the headers were out. So since the queue moved under /data in
1.3.0. Confirmed on the test server: selfpost-backup as root makes the
whole archive, the same command as user panel fails the same way.
Owner, 2026-10-11: the panel's archive comes without the queue.
- backup.Params.Unprivileged: the queue is left out, and so is a file
under certs/ that cannot be opened (the private key on the test
server is root and Postfix's group only), named through OnWarn. The
selfpost-backup command leaves it unset and takes both, as before.
- backup.Check reads through everything Create would archive, writing
nothing. The panel runs it before the first byte: a failure is a
message on the page, not a download cut short.
- The box, Help (design commit before this one) and the guide say what
the panel's archive leaves out; the guide's comparison table has rows
for the queue and the private key, and tells to put the key back
before starting a restored instance.
go vet ./internal/backup/ ./internal/web/handlers/: ok
go test ./internal/backup/ ./cmd/... ./internal/web/...: ok
make e2e on the test server, with the restore scenario: ok
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Owner, 2026-10-11: the archive made from the panel comes without the Postfix queue (variant 1). The Full backup box and the Help topic say so, and name the other thing the panel may not read - a certificate file, usually the private key - and the command that takes both.
check.py: 0 violations; outlines.json unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Until now the suite only read back the configuration generated for the
two optional port-25 features. DMARC ingest shipped in 1.7.0 with a
delivery service Postfix could not start, and nothing sent a report
through it until a real one arrived on the test server on 2026-10-11.
The last steps recreate the container with INBOUND_RELAY_ENABLE and
DMARC_REPORTS_ENABLE on, over the same /data, and talk to port 25 as a
stranger's mail server:
- the container log has no postconf warning, the session survives;
- a domain takes the SelfPost-hosted report address; a gzip report
mailed to it appears under Outbound > DMARC reports; an address nobody
chose is refused at RCPT;
- a report on another domain and a message with no report are accepted,
dropped with "dropped: ..." on their delivery lines, counted as two
parse failures, absent from the queue and from the reports;
- an inbound domain with the sink as upstream and one listed recipient:
mail for it reaches the sink, another recipient and another domain are
refused at RCPT.
The stage directory now holds docker-compose.yml and .env, as an
operator's project directory does (a full backup needs them; the
restore scenario that uses it is not in this commit - the panel's full
backup turned out not to work, see the next one).
make e2e on the test server: ok.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Security review of the DMARC report path (Fable, 2026-10-11), which ran
for the first time in 2.0.0. No Critical or High finding. Fixed:
- Medium: a report was bounded only as 8 MiB of XML - some 160 000
records or one name of megabytes, all of it rows and table cells. At
most 5 000 records (more is ErrRejected, dropped), text fields cut to
255 bytes on a character boundary, a record's count capped.
- Medium: the cap of 500 reports was for the whole server, so a stranger
filing under one domain's address evicted every other domain's
reports. PruneDMARCReports counts per domain.
- Low: the pipe service ran up to default_process_limit (100) commands
at once against the panel's database; maxproc is 1.
- Low: the reason logged for a dropped message carried the sender's
bytes (header line, zip entry name) as they came; cut to 200 bytes and
reduced to printable ASCII, zip entry names quoted.
- Low: SELFPOST_DATA_DIR / SELFPOST_DB_PATH with whitespace or "$" would
reach the command through export_environment as another path; refused
at start when DMARC ingest is on.
Left as they are, told to the owner: the parse-failure count is a
lifetime figure shown under "This week"; a report larger than the
command reads is dropped as unreadable; nested multipart is not searched.
The hub's Retention line follows the design commit before this one.
security.md: the review, and the accepted-risk entry says what a
stranger can and cannot do now. The roadmap no longer calls the review
pending.
go vet ./cmd/... ./internal/dmarc/ ./internal/store/: ok
go test ./cmd/... ./internal/dmarc/ ./internal/store/ ./internal/web/...: ok
make e2e on the test server (with the port-25 scenarios of the next
commit): ok
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The retention cap is counted per domain (the security commit after this one); the Ingest box says so: 500 reports a domain or 90 days.
check.py: 0 violations; outlines.json unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Owner, 2026-10-11: foreign and malformed reports are turned away.
The command answered every failure with EX_TEMPFAIL, so Postfix kept a
message that could never be stored and offered it again for days, each
attempt adding to the parse-failure count. IngestMessage now marks those
failures (unreadable as an aggregate report, a domain not sent from
here, another domain's report address) with ErrRejected; the command
counts one failure, writes the reason where Postfix puts it on the delivery line of the mail log
and reports the delivery as done. No bounce: the envelope sender of such
a message is whatever its author wrote, and the report addresses are
public. A failing store is still EX_TEMPFAIL.
go vet ./cmd/... ./internal/dmarc/: ok
go test ./cmd/panel/ ./internal/dmarc/ ./internal/web/...: ok
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The panel redesign (docs/plans/panel-redesign.md, 27/27). A breaking
release: every panel URL changed and the database starts from a single
baseline - a 1.x /data is not upgraded.
Since the last cut also fixed, all found on the test server:
- the panel can read the mail queue;
- port 25 answers with DMARC report ingest on, also next to the
inbound relay;
- DMARC reports reach the panel (the delivery service never worked
since 1.7.0);
- Postfix start-up warnings about the queue directory.
Cut: CHANGELOG [Unreleased] -> [2.0.0]; image pin 2.0.0 in compose,
README, guide; roadmap, plan and version table.
On the test server, on 252e5d1: go vet ./... ok, go test ./... ok
(22 packages), make e2e ok.
Full detail: CHANGELOG.md (2.0.0).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The first DMARC report to reach the test server (Google, 2026-10-11
09:41 UTC) was accepted on port 25 and then sat in the queue:
postfix/pipe: fatal: missing user= command-line attribute
postfix/qmgr: warning: private/dmarc-ingest socket: malformed response
Three things stood between an accepted report and the panel, all since
the feature shipped in 1.7.0 - no report had ever been delivered:
- The service was addressed as dmarc-ingest/unix-pipe. Its type is unix:
postconf -P answered "unmatched service_name/type" for flags=, user=
and argv= (the warnings in the container log at every start), which
are arguments of pipe(8) anyway and not -o parameters, and the service
stayed a bare "pipe". For the same reason postconf -MX never removed
it. It is dmarc-ingest/unix now, with the arguments in the service line.
- pipe(8) gives the command export_environment and nothing else, so the
command would have refused to run: it reads DMARC_REPORTS_ENABLE (and
the data directory, when it is not the default) from the environment.
Those are now named in export_environment.
- The command took the recipient from RECIPIENT, which local(8) sets and
pipe(8) does not. With no recipient the check that a report came to the
address of its own domain is skipped. It is now ${recipient} in argv,
one recipient per run.
When DMARC ingest is off the service and both parameters are removed.
e2e: every port-25 combination now asks Postfix for the service and the
two parameters (postconf -M / -h), and for their absence when ingest is
off - which is how the wrong type was found. make e2e on the test
server: ok. go test ./cmd/panel/ ./internal/dmarc/ ./internal/postfix/:
ok.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Follows the design commit before it: Machine, Processes, TLS certificate, Milter sockets and Reverse DNS lead to /server/health as it opens; so does the link in the Needs attention line, which takes its address from the card.
go vet ./internal/web/...: ok
go test -count=1 ./internal/web/...: ok
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Owner, 2026-10-11, after looking at the test server: the anchors work but mean little, Health is short and there is nothing to scroll past. The five cards (Machine, Processes, TLS certificate, Milter sockets, Reverse DNS) lose their #machine ... #hostname, as the links into Domain settings lost theirs.
check.py: 0 violations; outlines.json unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
/preflight is /server/preflight and the retention setting is on /server/settings; the routes are renamed, so the plan no longer says once they are.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The recursive chown to postfix at the end of the entrypoint also takes the queue directory itself and pid/, which Postfix keeps owned by root; postfix-script logs warning: not owned by root for both at every start. They are handed back to root after the chown, as public/ and maildrop/ are handed back to postdrop.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
initShowWhen bound a change listener to each control it found at DOMContentLoaded, so a form swapped in later would have been neither synced nor wired. The rules now listen once on the document and are applied again on htmx:load. No form arrives by a swap today; the roadmap row that recorded the gap is struck.
Checked in a browser on the application form: the rate-limit select toggles the manual and auto fields as before, and a copy of the form put into the page with htmx.swap is synced on arrival and follows its select. go vet and go test ./internal/web/... pass.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- The hint under "Last report: None yet" still said "Point rua= at the
hosted address below" - there is no such address any more, and nothing
is below. It names the three steps of the three-choice scheme: set a
domain to SelfPost hosted in its Domain settings, publish the DMARC
record the domain then shows, publish MX for the server's name.
- The hosted address form is Fact.Small, as in the mockup.
go test -count=1 ./internal/web/...: ok
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
On the test server the form dmarc-reports+<domain>@selfpost.mixfed.ru
broke over three lines in its column (2026-10-10). It takes the
supporting size the report id and rua of a report already use.
check.py: 0 violations.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Found by talking SMTP to port 25 of the test server once it answered
again (2026-10-10): RCPT TO:<dmarc-reports+bs.mixdep.ru@selfpost.mixfed.ru>
was refused with "550 5.1.1 User unknown in relay recipient table".
With INBOUND_RELAY_ENABLE and DMARC_REPORTS_ENABLE both true the server's
own name is a relay domain (for the reports) and relay_recipient_maps
lists the relay's recipients only. smtpd_reject_unlisted_recipient
applies after the restriction list, so the OK that
check_recipient_access gives a report address does not help: every
report was refused. The table of report addresses is added to
relay_recipient_maps in that combination.
Checked on the server by setting the parameter in the running container
and reloading Postfix: the domain's report address is answered 250, the
former server-wide mailbox and an unknown local part 550, a foreign
domain and an inbound domain without recipients 554. e2e: the combined
case requires relay_recipient_maps to name both tables.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Seen in the System log of the test server (2026-10-10): on every
connection to port 25
postfix/smtpd: fatal: unexpected command-line argument:
texthash:/data/postfix/dmarc_recipients,reject_unauth_destination,...
postfix/master: warning: .../smtpd: bad command startup -- throttling
With DMARC_REPORTS_ENABLE=true the script wrote
-o smtpd_recipient_restrictions=check_recipient_access texthash:...,...
into master.cf. Master splits a service line on whitespace, so smtpd got
the option cut at the space plus a stray argument and died at start:
port 25 greeted nobody - no inbound relay, no DMARC reports. postconf -P
smtp/inet/smtpd_recipient_restrictions showed the cut value,
"check_recipient_access". The server's logs have it under the previous
image too (2026-10-09), so it is not from the redesign; postfix check
does not notice, and the e2e check compared the text of master.cf with
an expectation that had the space in it.
- The restriction list is written with commas only (Postfix splits it
on commas as well as whitespace).
- master_cf_set_smtp_inet_o stops the container on a value with
whitespace instead of writing it.
- e2e: for every option it checks, the port-25 case also asks
postconf -P and requires Postfix to read the same value.
Not fixed here: Postfix's start-time warnings "not owned by root" for
the queue directory and its pid/ - the entrypoint's recursive chown
again; harmless, logged at every start.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Follows the design commit before it: the domain page's menu entries
Report address / Rate limit / Export, the DMARC record's note and the
delete page's "Export the domain" lead to /outbound/domains/{id}/settings
as it opens, not scrolled to a box.
go test -count=1 ./internal/web/...: ok
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Seen by the owner on the test server (2026-10-10): following Report
address, Rate limit or Export from the domain's menu opened Domain
settings scrolled down to that box. The page is one screen and a bit, so
the anchor only pushed the title and the first row out of view. The
links (the domain menu, the DMARC record's note, the delete page, Server
> Settings) lose their #reports / #limit / #export.
check.py: 0 violations; outlines.json unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The owner's choice (2026-10-10) for the last line of the Processes table
on Server > Health: postfix-reload runs when the configuration is
reloaded and exits, so between runs supervisord calls it STOPPED - a
word that reads as a fault beside six running programs, in a box that
rightly says ok. Its row now has the neutral tag "on demand" and says
when it last ran (or that it has not run since the container started).
A one-shot in trouble keeps supervisord's word and level, and a daemon
that is stopped is still an error.
go test -count=1 ./internal/web/...: ok
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Seen by the owner on the test server, on real data (2026-10-10).
- Overview's Processes card said "7 of 7 running" with six programs
running and postfix-reload idle. The card counts the programs that
are meant to stay up; a one-shot helper (health.Process.OneShot) is
left out of both numbers and named only when it is in trouble.
- The Milter sockets card clipped its second line ("OpenDKIM,
send-log,...") once the spam filter was the third socket. The line
names the sockets that do not answer when some do not; with all
answering it is their names, or "incl. the spam filter" - the
mockup's words - when the filter is among them.
- A real 2048-bit DKIM value did not fit the three rows of its field
and scrolled inside it. The field is as tall as the value needs
(3 to 12 rows).
The fixtures had a shortened key, two sockets and no one-shot program,
so none of this showed in the evidence screenshots.
go test -count=1 ./internal/health/... ./internal/web/...: ok
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The entrypoint fix of 783bca8, with what the two e2e runs on the test
server showed: the new subtest fails on the commit before the fix
("the panel could not run postqueue -p") and the whole suite passes on
the commit with it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The entrypoint ends with chown -R postfix:postfix on the queue tree, on
every start, after postfix set-permissions has run. That puts public/
and maildrop/ in group postfix; Postfix keeps them in its setgid group,
and postqueue (setgid postdrop) reaches the showq socket through that
group only. So the panel, which runs postqueue -p as its own user, got
"Connect to the Postfix showq service: Permission denied": the Queue
card of Overview said "Could not be read" and the queue page showed no
listing. The data directory of the test server has carried the wrong
group since 2026-08-18, so this is older than the redesign - the
recursive chown came with the persistent queue in 1.3.0; Postfix itself
logs "warning: not owned by group postdrop" for both at start.
The two directories are handed back to postdrop after the chown.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A subtest that opens Outbound > Queue right after sign-in and expects
postqueue's own line for an empty queue, not the page's "Could not read
the mail queue". Seen on the test server after the 2.0 deployment: the
Queue card said "Could not be read", and the panel's log had
"postqueue: fatal: Connect to the Postfix showq service: Permission
denied" on every poll. The fix follows in the next commit; this one is
expected to fail on its own.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Follows the design commit before it.
- internal/dmarc: the ingest allow-list is the report address of each
domain that is on this host and nothing else;
DefaultHostedReportAddress and DefaultHostedSuggestion are removed.
Since 1.7.0 the general mailbox dmarc-reports@<hostname> was open
while no panel-wide default existed; the three-choice scheme had
left it open always.
- The DMARC hub's Ingest box shows the form of a hosted address,
dmarc-reports+<domain>@<hostname>.
- guide.md: what port 25 accepts.
Evidence: docs/assets/panel-redesign/evidence/dmarc.png re-shot (the
address form wraps after the plus sign in its column at 1440 px).
go test -count=1 ./internal/dmarc/... ./internal/web/... ./cmd/panel: ok
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The owner's decision (2026-10-10): the mailbox for the server as a
whole, dmarc-reports@<hostname>, goes. Under the three-choice scheme a
domain can only choose its own hosted address, so nothing can be told
to report to the general one. The Ingest box's fact becomes "Hosted
addresses: dmarc-reports+<domain>@<hostname>".
check.py: 0 violations; outlines.json unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The owner's instruction (2026-10-10). panel.js sets is-hidden, beside
the hidden attribute, on a block that a control switches off - the
fields of a mode that is not chosen, the password fields of an unticked
encryption box, the domain lists of a global user - because hidden alone
does not hide an element whose stylesheet gives it a display of its own
(a box, a grid row). No screen writes the class and it draws nothing, so
the components page has nothing to show for it; the list now names
everything the panel uses.
check.py: 0 violations.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Stage 3, step 1 (Opus). Each step of stage 2 was reviewed against its
mockup when it landed; this closes the review over the whole, after the
changes of stage 3 (Help topics and the ? on boxes, the three-choice
report address, row actions, the filter's socket).
- Evidence: all 29 screenshots under docs/assets/panel-redesign/
evidence/ are re-shot from the current templates - the 28 pages of
pageFixtures and the sign-in page without an administrator - 1440 px
wide at each page's full height. Headless Edge against the real
templates served locally with their fixtures; NOT from the container
(no Docker on this machine) - the container is checked on the test
server at deployment.
- panel/audit.js over the same 29 pages at 1440 and 1280: 4160 texts
checked, problems: [] (type floor, contrast, sideways scroll,
overflow in boxes and heads, scrolling tables, clipped card values,
missing icons). Every ? link has the form /help#topic, and
TestEveryHelpLinkLandsOnATopic proves each lands.
- Guards: outlines, class vocabulary, page structure, components-only,
stylesheet equality, routes, static allow-list - green; the ratchet
is empty.
Where the product knowingly differs from a mockup (each recorded in the
commit of its step): the queue is the postqueue listing, not a table;
the outbound log has no status filter; the message history has no
"accepted from IP" step; Health draws the sockets this server has;
the Spam filter box states only what the configuration says; the user
form's e-mail help promises nothing that is not built; Overview's title
does not claim that mail is flowing. Roadmap screens (in-log, in-queue,
in-quarantine, in-filter-lists, preflight) are drawn and not built.
Open with the owner: is-hidden, which panel.js sets at run time, is not
in check.py's Bulma subset (the edit of check.py is his).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Stage 3, step 3. Written by Sonnet from the code and the plan, checked
by Opus.
- guide.md: a section on the panel - menu, paths, the roles global and
domain with outbound and inbound assigned separately -, Overview and
Health, Queue and System log, Account and Server > Settings, the
domain page / Domain settings / application form, Help as a page of
topics. The DMARC report address is one of three choices set in
Domain settings. The Spam filter box, the two INBOUND_ANTISPAM_*
variables the panel reads, and the filter's socket on Health.
Sessions; backup and export encrypted unless the box is unticked;
1.x data is not restored into 2.0.
- architecture.md: Panel HTTP surface rewritten from web.go (routes,
the /server/ guard, per-object checks, <page>/fragment, one layout,
typed page data, the kit and its guards, CSP and htmx); Persistence
gains the database - one baseline, a 1.x file refused, the tables as
they are.
- security.md: roles and what each reaches, as built; sessions by user
id and what ends them; state-changing requests; errors shown to
users; the 2.0 review; two accepted risks from it - a delegated
inbound upstream may be any host and port, and Account's profile is
saved without the password.
- CHANGELOG: Removed opens with "2.0 starts from an empty data
directory"; entries that describe the user-default report address
say they were replaced before release; the filter's socket.
- roadmap.md: two review follow-up rows struck as retired by the
redesign (.pair / .split, the two back links); password-reset notes
that changing the account e-mail must then ask for the password.
Progress 25/27.
- development.md: the component kit and its guards; README and
product.md: wording.
Left as they are
- The third follow-up row the plan expected to strike - initShowWhen
is not re-run after an htmx swap - still holds: no polled fragment
carries such a control today, so nothing breaks, but the redesign did
not retire it.
- guide.md embeds no screenshots, so none needed retaking.
- README and guide still name image 1.9.5 - the version cut.
- plans/preflight.md and plans/inbound-antispam-panel.md use pre-2.0
paths; roadmap's template-data-typing is largely done by the
redesign.
go test -count=1 ./cmd/panel (env-doc tables): ok
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The owner's decision of 2026-10-10, as the plan has it (Who sees what:
for the global administrator the filter's socket is one of the milter
sockets on Health and Overview). Written by Sonnet, reviewed by Opus.
- internal/health: CheckMilter probes an address in Postfix's two
forms. unix:/path is the existing stat; inet:host:port is a TCP
dial with a 1 s cap, closed at once, nothing written. An address in
neither form is reported as not checkable, never as fine. Details
are plain sentences ("No answer from antispam:11332 within 1 s.",
"Connection to ... refused.", "The name ... could not be
resolved.") and never the Go error text.
- Health: with the filter on, a third row "Spam filter - inbound"
with the configured address; it counts in the socket status and in
Overview's Milter sockets card. A filter that does not answer is an
error when INBOUND_ANTISPAM_MILTER_ACTION is tempfail (inbound mail
is deferred) and a warning otherwise (mail goes through
unfiltered), and the row says which. With the filter off nothing
changes.
- Both pages are global-only; a test checks a domain administrator
never receives the address.
The dial runs with the other checks on every request and poll of
Overview and Health (5 s active, 30 s idle, none while the tab is
hidden): a dead filter costs a poll at most 1 s, and a live one sees a
connect-and-close at that rhythm per open tab.
Not changed: fixtures and screenshots (the mockup draws two sockets),
help texts. The CHANGELOG line comes with the documentation step, which
is editing that file now.
go test -count=1 ./internal/health/... ./internal/web/... ./cmd/panel
ok (seven packages)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Follows the design and guard commits before it. Written by Sonnet,
reviewed by Opus.
The report address
- Baseline: users.dmarc_default_mode, users.dmarc_default_address and
domains.dmarc_rua_user_id are gone. domains.dmarc_rua alone says
where reports go: '' is no reports (a new domain), the address
SelfPost derives for the domain is the hosted choice (while hosted
reports are on), anything else is a typed address.
- Removed: the user default and its store functions, a domain
following a user, keep / inherit, followCreator, POST /account/dmarc
with its handler, the DMARC box of Account, the user-delete item
about domains losing their address.
- Domain settings offers the three choices; beside the typed address a
"Use my e-mail" button fills in the profile e-mail (panel.js: one
generic data-fill hook; absent without a profile e-mail; it fills,
it does not save).
- The DMARC record's note names the address and no user. This closes
L4 of the security review.
- Domain export / import carries the address as a plain string.
- The ingest allow-list is the generic mailbox plus each domain's
address on this host; the generic mailbox is always open now.
- dev/migrate1x (not in the tree): a 1.x domain that inherited the
instance default gets that address as its own, so no published rua=
changes.
Row actions
- row_action partial (RowAction: Label, Post, Confirm, Danger);
panel.css has the mockup's three rules. A row of the Applications
table is Edit, New password, Delete; the confirmation texts are the
ones the application page's head uses. Both roles, with the checks
of the application page.
Tests cover: a new domain has no reports; each choice saved and shown
selected; hosted refused when hosted reports are off; a bad address
refused with what was typed kept; an untouched save changes nothing in
each state; the fill button with and without a profile e-mail; deleting
a user leaves every domain's address; the row's forms, targets and
confirmations for both roles and a 404 for a stranger.
Known: if hosted reports are switched off later, a stored hosted address
reads as a typed one. docs/guide.md still describes the old choices -
the documentation step.
Evidence
- docs/assets/panel-redesign/evidence/account.png,
out-domain-settings.png, out-domain.png, user-delete.png, help.png:
1440 px wide, headless Edge, real templates served locally with
their fixtures - NOT from the container.
- panel/audit.js on out-domain, out-domain-settings, account, help,
components and user-delete at 1440 and 1280: 1420 texts checked,
problems: [].
- In a browser: the fill button sets the field and does not submit;
the row buttons have the font, size and colour of the Edit link,
Delete the red of the other delete links; data-confirm fires on
both. Hover and focus states were not observed.
- go test -count=1 ./internal/web/... ./internal/store/...
./internal/dmarc/... ./internal/app/... ./internal/inbound/...
./cmd/panel: ok (ten packages; dev/migrate1x also ok).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The owner's decision of 2026-10-10 (the design commit before this one)
removes the per-user default report address, and with it the form that
POST /account/dmarc served. The plan's Routes table no longer lists it;
the guard's copy of that table follows. Nothing else in the guard
changes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Two decisions of the owner, 2026-10-10.
The report address. "Three choices: no reports, the SelfPost address, a
typed address. No reports by default. A button that fills the typed
address with one's own e-mail from the profile." This replaces the
2026-09-21 design in which a domain followed a named user's default: a
co-administrator of a domain was shown another user's login name, and
three explicit choices are simpler.
- src/out-domain-settings.html: the box is a select of the three
choices and the address field with a "Use my e-mail" button.
- src/out-domain.html: the DMARC record's note names the address, with
no user in it.
- src/account.html: the DMARC reports box is gone; Account is Profile
and Password, one box per column.
- src/settings.html: the lead no longer points at Account for the
report address.
- docs/plans/panel-redesign.md: Account e-mail rewritten, the Routes
row of account without /account/dmarc. docs/schema-migrations.md:
the columns that go.
Row actions. New password and Delete in a row of the Applications table
are POSTs, so each is a form with a button that looks like the links
beside it - as Sign out became one.
- src/out-domain.html: Edit / New password / Delete per row, the box
foot about confirmations is back; the login cell does not wrap.
- theme.css: .table td.sp-actions form sits inline with the links'
spacing, and its button takes the link's colour and type.
- src/components.html shows the pattern once.
check.py: 0 violations. outlines.json: only account changed (two equal
columns, one box each).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Follows the design commit before it. Written by Sonnet, reviewed by
Opus.
- view/help.go, templates/help.html: the mockup's topics, texts, ids
and menu. The page shows the groups the reader's menu shows:
Overview and Server for the global role, Outbound and Inbound by
reach (Inbound only with the feature on), Account for everyone;
boxes are numbered as shown.
- Box.Help set on the boxes the mockups mark; the inbound MX box has
its link again.
- TestEveryHelpLinkLandsOnATopic also fails on a topic that no page
links to, unless it is listed as menu-only (account) - and on a
stale exception.
Known until the change that follows: the dns, dmarc and account topics
describe the three-choice report address, which the code does not have
yet; Account's DMARC box still links to #dns.
Not verified by running: three sentences rest on Postfix behaviour
(delivery is deferred when a required TLS session cannot be made; a
milter with default_action=tempfail answers 4xx; with "any recipient"
an address the upstream refuses comes back as a bounce).
Evidence
- docs/assets/panel-redesign/evidence/help.png (1440x3600) and
in-domain.png: headless Edge, real templates served locally with
their fixtures - NOT from the container.
- go test -count=1 ./internal/web/... ./cmd/panel: ok (six packages).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The owner's decision (2026-10-10): help is needed by almost every
section, not by Overview and Outbound alone. The Help mockup had six
topics while other screens already linked to topics it did not have
(#dmarc, #inbound).
- src/help.html: seventeen topics in five boxes - Overview (checks,
certificate, rdns), Outbound (dns, apps, limits, log, queue, dmarc),
Inbound (inbound, upstream, recipients, filter), Server (backup,
users, settings), Account (account) - one paragraph each, and the
side menu lists them by group.
- A box head gets the sp-help link where a topic now exists: health,
out-domain-settings, out-app, out-log, out-message, out-queue, the
three DMARC screens, in-domains, in-domain, backup, users, user,
settings. No link on boxes that explain themselves (confirmations,
the shown-once password, sign-in) and at most one per box.
- outlines.json regenerated (check.py --write): only help changed -
Outbound holds six topics, and Inbound, Server and Account are new
boxes. 0 violations.
The dns, dmarc and account topics describe the report-address scheme the
owner decided the same day (no reports / the SelfPost address / a typed
address, no user default); the account and out-domain-settings mockups
are redrawn for it in the change that follows.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Stage 3, step 2. Reviewed by Fable (read-only), fixed by Sonnet with a
failing test before each fix, checked by Opus. No Critical or High
finding: route authorization, tenant isolation by id, the /server/
subtree guard, output encoding and the origin check were found sound.
Fixed
- M1: a session was keyed by user name and had no link to the user. A
deleted or renamed user's cookie stayed alive (renewed on every hit,
before the user was looked up) and became a session of whoever was
later created under that name, with that account's role. The
baseline now has sessions.user_id REFERENCES users(id) ON DELETE
CASCADE; the principal is resolved by id and a session is renewed
only after its user was found. RenameSession is gone.
- M2: any user's password change deleted every other user's sessions.
It ends that user's other sessions only.
- L1: a password set on the user form ends that user's sessions (an
administrator editing their own account keeps the current one).
- L2: store, saslpasswd2 and Postfix reload errors were shown verbatim
on the application form, the inbound upstream / recipients forms and
the limit recalculation - pages a domain administrator reaches.
Validation errors (app.ValidationError, inbound.ValidationError,
store.ErrNotAutoRateLimit) keep their text; anything else is logged
and shown as "Could not save ... Check the logs and try again."
- L3: store.ListDMARCReports treated an empty domain list as "all". It
takes a scope now (AllDomains or a list; an empty list is no rows),
and the DMARC hub is open to the domain role as the plan says: the
reports of the assigned domains, without the instance-wide ingest
box. The menu entry was already shown to them and led to a 404.
- I1: /static/ answered with a listing of the embedded assets. A path
that is not a known asset is a 404.
Consequences
- The global DMARC hub lists reports of domains that were deleted
since, too.
- dev/migrate1x (not in the tree) no longer carries sessions over:
everyone signs in again after the transfer.
- docs/schema-migrations.md: the sessions row.
Recorded, not changed - for the owner:
- L4: the report-address select names the followed user and their
address to every administrator of the domain, by default the global
administrator's login name. The plan asks for it.
- L5: a delegated inbound upstream may be any host and port, loopback
and private ranges included - within the trusted-operator model; to
be listed among the accepted risks in security.md.
- L6: Account's profile is saved without the password; becomes a
requirement once password-reset mails to users.email.
- Still echoing an error's text: backup and domain import (global
only), the domain form's validation messages.
Not checked by the review: anything needing the container (e2e, real
Postfix / DNS), CSP behaviour in a live browser.
go test -count=1 ./internal/web/... ./internal/store/... ./internal/app/...
./internal/inbound/... ./internal/dmarc/... ./cmd/panel
ok (all eleven packages; dev/migrate1x also ok)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Stage 2, steps 7 and 8 in one commit: the static guard fails as soon as
the ratchet holds no page while legacy.css still exists, and the tests
of the old layout need it while that layout exists - the two steps do
not split into states that pass on their own. Written by Sonnet,
reviewed by Opus.
The queue
- out-queue, named after its mockup: today's content - the listing of
postqueue -p in a log pane, polled as before (GET
/outbound/queue/fragment) - and the "How retries work" facts. The
mockup's table of messages is the roadmap's queue-as-a-table; the
outline guard exempts this page for exactly that reason. Routes and
the global-only check are unchanged; the page has no actions and
never had.
- Inbound has no Queue entry and no route serves one.
The old design is gone
- Deleted: layout_legacy.html, legacy.css, logo-compact.svg,
mail_queue.html, mail_queue_body.html.
- view.go has one path: every page is layout.html + components.html +
its own files; kitPages, kitFragments, the map-data branch of Render,
the dict and back template functions and metaOf's map branch are
removed. handlers: pageBase removed.
- legacy_pages.txt holds its header and no entry.
- panel.js: the old copy handler (button.copy in .code-row) is
removed. A show-when rule now applies to every element it matches,
and hiding adds Bulma's is-hidden beside the hidden attribute - a
.box or a grid row has a display of its own and stayed visible.
- The user form hides both domain lists again when the role is Global
(Box.DomainPick, an attribute; data-global-role on the select).
- Tests of the old layout are deleted; what they asserted about the
shell (version only signed in, the menu follows the user's reach,
Inbound only with the feature, Help present, the current entry
marked) is asserted on the kit pages, two tests added for it.
For the owner: is-hidden is a Bulma class outside the subset of
check.py. No template writes it - the script sets it at run time to
hide a block - so no guard sees it; say if it should be on the list.
Evidence
- docs/assets/panel-redesign/evidence/out-queue.png: 1440 px wide,
headless Edge, the real template served locally with its fixture -
NOT from the container (no Docker on this machine).
- panel/audit.js on out-queue, user and backup at 1440 and 1280:
366 texts checked, problems: [].
- In a browser against the local fixture server: the encryption
fields of Backup, the manual / auto limit fields and the custom
report address of Domain settings show and hide with their control;
all 28 fixture pages answer 200 with panel.css, /static/legacy.css
answers 404.
- Not exercised: polling against a real postqueue (none on Windows).
- go test -count=1 ./internal/web/... ./cmd/panel
ok internal/web
ok internal/web/auth
ok internal/web/handlers
ok internal/web/validate
ok internal/web/view
ok cmd/panel
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Stage 2, step 6. Written by Sonnet, reviewed by Opus.
- Six pages named after their mockups: system-log, backup, users,
user, user-delete, help, with typed page data (view/server.go,
help.go), fixtures and tests. The old templates of the six are gone
(user_form.html, encrypt_fields.html deleted, the others rewritten)
and their names are off legacy_pages.txt, which now holds mail_queue
alone.
- Help is six help_topic blocks (checks, certificate, rdns, dns, apps,
limits); a test renders every fixture page and fails on a Help link
without a topic. The help drawer (help_drawer.html and its includes
in layout_legacy.html) is removed; its rules in legacy.css go with
the next step.
- The user form sets the user's e-mail (plan, Account e-mail): create
and update, validated, and a changed address resyncs DMARC for a
user whose default is the account e-mail - through the helper the
Account profile now shares.
- Handlers pass typed data; routes, POST targets, field names,
validation and authorization are unchanged. A test loads the user
form for four reach combinations and for the only global user, posts
it back untouched and finds the user unchanged.
Fixed on the way
- The user delete confirmation said the domains were not affected. A
domain that follows the user's default report address is left with
none; the page names those domains now.
- A refused backup lost the encryption choice and printed a literal
<code> element in its message.
Differences from the mockups, for the owner:
- Help has no DMARC and no Inbound topic: its outline fixes two boxes
of three topics, although other mockups link to #dmarc and #inbound.
Until the owner decides, Account's Help link goes to #dns and the
inbound MX box has none;
- system log: newest line first, as drawn; "refreshes on its own" (5 s
active, 30 s idle); only status=, warning: and error: lines are
coloured;
- backup: the file field is a plain file input; the encryption box is
ticked, as drawn and as on the domain export, so a plain download
means unticking it first;
- users: an E-mail column; the flash does not name the user;
- user form: no lead; the domain lists are no longer hidden by script
when Global is chosen (the server ignores them for that role).
Evidence
- docs/assets/panel-redesign/evidence/system-log.png, backup.png,
users.png, user.png, user-delete.png, help.png: 1440 px wide,
headless Edge, real templates served locally with their fixtures -
NOT from the container (no Docker on this machine).
- panel/audit.js on the six pages at 1440 and 1280: 720 texts checked,
problems: [].
- The system log's polling swap was not exercised in a browser; it
uses the mechanism of Overview and Health.
- go test -count=1 ./internal/web/... ./cmd/panel
ok internal/web
ok internal/web/auth
ok internal/web/handlers
ok internal/web/validate
ok internal/web/view
ok cmd/panel
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Stage 2, step 5. Written by Sonnet, reviewed by Opus.
- Three pages named after their mockups: in-domains, in-domain,
in-domain-delete, with typed page data (view/inbound.go), fixtures
and tests. The engine pages inbound, inbound_domain and
inbound_delete are gone with their templates and are off
legacy_pages.txt.
- The Spam filter box (plan, Who sees what). The panel now reads
INBOUND_ANTISPAM_MILTER and INBOUND_ANTISPAM_MILTER_ACTION, with the
semantics of build/postfix-config.sh (on only with the inbound relay
on and a milter set; accept | tempfail). It reports them, nothing
more:
on Applies to / Filter (the address, global role only) / If the
filter is down (accepted unfiltered | deferred)
off one sentence; the variable is named for the global role only
The list shows "Spam filter on|off" in its head.
- Handlers pass typed data; routes, POST targets, field names,
validation and authorization (global: all inbound domains; domain
role: the assigned ones, another tenant's id is a 404; add and
delete are global-only) are unchanged. A test saves the upstream and
recipients forms untouched and finds the domain unchanged.
Left out of the mockups' Spam filter box and menu - no source before
inbound-antispam-panel (2.1): the engine name, the quarantine row, the
"refused at the door, see the log" row, the lists row, "Log for this
domain". Also "Checked N min ago" on the MX box: the check carries no
time.
Not done: the filter's socket on Health and Overview. internal/health
checks a unix socket by stat; the documented form of the filter's
address is inet:host:port and needs a dial that does not exist yet.
Differences from the mockups, for the owner:
- refusals and confirmations are the kit flash above the boxes;
- the MX tag is coloured by the check in every state; an MX that
points elsewhere shows "In DNS now";
- the lead of the domain page follows what is saved ("Not accepting
mail yet: no upstream is saved");
- the delete page lists only what the domain has;
- /help#inbound opens Help at the top until the Help step.
Evidence
- docs/assets/panel-redesign/evidence/in-domains.png, in-domain.png,
in-domain-delete.png: 1440 px wide, headless Edge, real templates
served locally with their fixtures (filter on) - NOT from the
container (no Docker on this machine).
- panel/audit.js on the three pages at 1440 and 1280: 352 texts
checked, problems: [].
- That the panel process sees the two variables inside the container
is read from supervisord.conf and entrypoint.sh, not run.
- go test -count=1 ./internal/web/... ./cmd/panel
ok internal/web
ok internal/web/auth
ok internal/web/handlers
ok internal/web/validate
ok internal/web/view
ok cmd/panel
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Stage 2, step 4. Written by Sonnet, reviewed by Opus.
- Five pages named after their mockups: out-log, out-message, dmarc,
dmarc-domain, dmarc-report, with typed page data
(view/outbound_log.go, dmarc.go), fixtures and tests. The engine
pages deliveries and delivery are gone with their templates; the
three DMARC templates are written anew. All five old names are off
legacy_pages.txt.
- The log polls its table (GET /outbound/log/fragment, as before); the
page count in the box foot follows out of band, and the filter in
the head is no longer swapped away while it is being chosen.
- Handlers pass typed data; routes, query parameters, filters, paging
and authorization are unchanged.
- Kit: Foot.ID / Foot.OOB (attributes only, as on Head) and Fact.Small,
the mockup's "sp-mono sp-small" value, for a report id and an rua
that wrapped mid-word in their column.
- test/e2e status scrapers match the new tag markup. Not run: no
Docker here.
Fixed on the way
- The DMARC domain page read a field its source rows do not have and
answered 500 as soon as the domain had a source. It builds the rows
from the store's rollup now, with a test.
Differences from the mockups, for the owner:
- the log has no Status filter (the store filters by domain and
application only); a refused message is "rejected", a neutral tag as
the mockup's "rate-limited" is;
- the message history keeps the panel's own wording with full dates and
has no "accepted from <IP>" step - the client address is not stored;
- dmarc: before the first report the box says "None yet" with what to
check; a quiet mailbox is an amber "quiet" tag;
- dmarc-domain: the stamp is pass / (pass + fail) rounded down, amber
on any failure, red when nothing passed, absent without mail;
- dmarc-report: the title has no year; "this relay" marks record rows
as well as sources.
Known and left as it is (not introduced here): the DMARC hub answers 404
to a domain administrator while the menu shows the entry and the plan
gives that role the reports of its own domains. Opening it needs the
report list to refuse an empty domain filter first - for the stage-3
security review.
Evidence
- docs/assets/panel-redesign/evidence/out-log.png, out-message.png,
dmarc.png, dmarc-domain.png, dmarc-report.png: 1440 px wide, headless
Edge, real templates served locally with their fixtures - NOT from
the container (no Docker on this machine).
- panel/audit.js on the five pages at 1440 and 1280: 784 texts checked,
problems: [].
- Polling of the log was exercised in a browser against a local
fixture server with document.hidden overridden: four polls, one
table and one foot after each, the filter untouched.
- go test -count=1 ./internal/web/... ./cmd/panel
ok internal/web
ok internal/web/auth
ok internal/web/handlers
ok internal/web/validate
ok internal/web/view
ok cmd/panel
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Stage 2, step 3. Written by Sonnet, reviewed by Opus.
- Six pages named after their mockups: out-domains, out-domain,
out-domain-settings, out-app, out-app-created, out-domain-delete,
with typed page data (view/outbound.go, outbound_app.go), fixtures
and tests. The engine pages dashboard, domain_detail and
domain_delete are gone with their templates and are off
legacy_pages.txt. GET .../settings has its own handler.
- Handlers pass typed data; routes, POST targets, field names,
validation, authorization (both roles, another tenant's id is a 404)
and Cache-Control: no-store on the password page are unchanged.
- Kit, attribute only: Action.Confirm writes data-confirm on a POST
action's form (panel.js already reads it).
- test/e2e/panel_client.go scrapes the new markup (copy fields, the
application row's Edit link). Not run: no Docker here.
Fixes found on the way
- Report address of a domain that follows another user's default: the
select had one "inherit" option, selected, so saving the form
untouched moved the domain to whoever saved it. It now offers that
user's default as "keep" (selected, saves nothing) and "My default
instead" as the change, as the mockup draws it; "keep" on a domain
that does not follow someone else is refused.
- panel.js: syncShowWhen stopped at the first matching rule, so the
auto-limit fields were never toggled with the limit mode. Every
matching rule applies now.
Differences from the mockups, for the owner:
- application rows carry Edit only. New password and Delete are POSTs,
and the kit has no link-styled POST control for a table row; both are
in the application page's head with a confirmation. Drawing them in
the row needs a design: change (as Sign out was);
- the application login is read-only on edit (the handler refuses a
rename); the submit says Create application on the empty form;
- multiplier bounds are the store's 1.5-5, default 2.5;
- "Recalculate" buttons exist where the stored limit is automatic;
- a domain administrator has no add box and no Delete; Delete domain is
global-only on the domain page and on settings;
- a fourth record, Report authorization, appears when the report
address is on another domain; record states missing / weak / not
checked are tags with the explanation under the value;
- refusals are the kit flash above the first box, and those of the
settings forms now show on the settings page; successful saves still
redirect to the domain page;
- the export checkbox is ticked by default, as drawn.
Evidence
- docs/assets/panel-redesign/evidence/out-domains.png, out-domain.png,
out-domain-settings.png, out-app.png, out-app-created.png,
out-domain-delete.png: 1440 px wide, headless Edge, real templates
served locally with their fixtures - NOT from the container (no
Docker on this machine). Taken before the report-address fix, which
does not change what the fixture renders.
- panel/audit.js on the six pages at 1440 and 1280 (text inside
[hidden] blocks skipped): 866 texts checked, problems: [].
- go test -count=1 ./internal/web/...
ok internal/web
ok internal/web/auth
ok internal/web/handlers
ok internal/web/validate
ok internal/web/view
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Stage 2, step 2. Written by Sonnet, reviewed by Opus.
- Four pages named after their mockups: overview, health, account,
settings, each with a typed page struct (view/overview.go, health.go,
account.go), a fixture and tests. The engine page "status" is gone
(status.html, status_body.html deleted); "settings" is Server >
Settings only. status and settings are off legacy_pages.txt.
- Overview: the stamp and its sentence are derived from the six cards
(the worst card sets the level and the lead names the cards that are
not fine); the outbound and inbound domain tables are built from the
existing service calls. Without the inbound relay the outbound box
takes the full width.
- Both pages refresh themselves as the old Status page did. Health
polls its first row of boxes and swaps the second out of band.
Overview polls its Server health box and the page head follows out of
band (GET /overview/fragment), so stamp, sentence and time move with
the cards.
- Kit, attributes only (no class, no CSS, markup unchanged when empty):
Box.Poll and Head.ID / Head.OOB.
- healthCards: the Queue card says "Empty" / "N queued" instead of
postqueue's line, which was clipped in a card; Processes and Milter
sockets carry a sub line. OverallStatus / OverallHeading are gone.
- Handlers pass typed data; routes, POST targets, field names,
validation and authorization are unchanged.
States the mockups do not draw, all from existing partials: errors and
successes as the kit flash above the boxes; Health with an unreadable
process list, a certificate without facts, a missing host name; Account
with the report authorization not published / not found / not checked
and with hosted reports off; empty domain tables as empty_state.
Differences from the mockups, for the owner:
- Overview title is one of "Everything is running normally." /
"Running, with warnings." / "A component needs attention.", not
"Mail is flowing..." (the checks do not prove that); the note says
"Updated HH:MM:SS UTC" without an interval, Health's lead says
"Refreshes automatically" (panel.js polls at 5 s active, 30 s idle);
- one Queue card ("outbound"), two milter sockets: what this build has;
- the days column is the stats window, min(30, retention);
- Health: no "check port 80" paragraph (the certificate's own detail
says what to check); a failing socket shows its detail under the path;
figures come from the sampler ("12%", "MiB");
- Account: the DMARC form keeps the address field for "Another
address" (hidden unless chosen); the count "N of M now" is the
domains following this user's default;
- Settings: Level 2 links the domain list only;
- /help#checks and /help#dmarc open Help at the top until the Help step.
Evidence
- docs/assets/panel-redesign/evidence/overview.png, health.png,
account.png, settings.png: 1440 px wide, full height, headless Edge,
real templates served locally with their fixtures - NOT from the
container (no Docker on this machine).
- panel/audit.js on the four pages at 1440 and 1280: 612 texts
checked, problems: [].
- Polling was exercised in a browser against a local fixture server
with document.hidden overridden (the pane was hidden): one element
per id after a swap, no leftover hx-swap-oob, about one request per
5 s. The first request of a visible tab was not observed.
- go test -count=1 ./internal/web/...
ok internal/web
ok internal/web/auth
ok internal/web/handlers
ok internal/web/validate
ok internal/web/view
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Event notifications are not built (panel-notifications is a roadmap
candidate), and the outline of both screens forced a box that could only
say so. The owner's decision (2026-10-10): take the boxes out of the
mockups; they are drawn again as the first step of that feature's plan.
- src/account.html: box 03 Notifications removed, DMARC reports is 03;
the lead and the e-mail help no longer promise notifications or a
reset link.
- src/settings.html: box 02 Notifications removed, Sending rate limits
is 02; the lead no longer lists notifications.
- outlines.json regenerated (check.py --write): the two boxes are gone
from account and settings, nothing else moved. 0 violations.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- panel.css: .sp-signin-form .sp-form, as theme.css has it now.
- templates/login.html: the state is div.sp-form with the mockup's
heading and line; the empty form of ef8a82e is gone.
- tests: the hint is checked by its new wording, and that the page
carries no form at all.
Evidence
- docs/assets/panel-redesign/evidence/login-no-admin.png: 1440x900,
headless Edge, real templates served locally (no Docker here).
login.png re-shot after the selector change: byte-identical.
- go test -count=1 ./internal/web/... : ok (web, auth, handlers,
validate, view)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The owner asked for the state to be drawn (2026-10-09) instead of the
form.sp-form without action or controls that ef8a82e used to get the
column centred.
- panel/src/login-no-admin.html: the signed-out screen with a
div.sp-form in place of the form - host kicker, the heading
No administrator yet, one muted line pointing at the one-time setup
link. No new class: div.sp-form is on the components page already.
- theme.css: the signed-out column is sized through .sp-form
(.sp-signin-form .sp-form), not through the form element, so the
column does not need a form to be centred. Same declarations.
- index.html: the screen is on the map; thirty-four screens.
check.py: 0 violations; outlines.json unchanged (bare screens have no
outline).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Stage 2, step 1. Written by Sonnet, reviewed by Opus.
- templates/login.html, setup.html: written from the mockups as the
content of layout_signed_out - form.sp-form, kicker, h1, Bulma fields
with icons, the full-width primary button. Actions, field names,
autocomplete and autofocus are the old ones.
- view/signin.go: typed Login and Setup (Meta carries a title only, so
Engine.Render takes the signed-out layout); auth handlers pass them
instead of maps.
- login and setup are in kitPages, have fixtures, and are off
legacy_pages.txt.
States the mockups do not draw:
- a refused sign-in / setup is the kit's flash (notification is-danger
is-light) above the fields;
- login before any administrator exists shows the old sentence in a
form.sp-form that has no action and no control: the signed-out
stylesheet sizes and centres its column through "form" alone. A
workaround, listed for the owner.
Differences from the mockups, for the owner:
- the kicker is <p><span class="sp-kicker"> (the kit's partial is a
span; a page may not write the class), the mockup has <p
class="sp-kicker">: the block sits about 3px lower;
- login's kicker is the configured host name and is absent when there
is none; setup's is the mockup's fixed text;
- the footer is the layout's legal line (full copyright holder name);
- the user name field has autofocus, so it shows its focus ring.
Evidence
- docs/assets/panel-redesign/evidence/login.png, setup.png: 1440x900,
headless Edge, from the real templates served locally with their
fixtures - NOT from the container (no Docker on this machine).
- panel/audit.js on login, login with an error, login without an
administrator, setup, setup with an error, at 1440 and 1280:
118 texts checked, problems: [] ; no page is taller than 900px.
- go test -count=1 ./internal/web/...
ok internal/web
ok internal/web/auth
ok internal/web/handlers
ok internal/web/validate
ok internal/web/view
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Stage 1, step 5 - and the end of stage 1.
- GET /server/health (+ /fragment) shows every check in full and is the
home of POST .../recheck and POST .../reload, which now return there.
- GET /overview (+ /fragment) shows the overall verdict and six cards -
machine, processes, TLS certificate, queue, milter sockets, reverse DNS -
each a view.HealthCard (the typed input of the kit's health_card) that
links to its table on Health. A check that could not run is a warning,
not "ok".
- Both pages read healthChecks(), so a card and its table cannot disagree.
Until stage 2 they share the old status template, switched by a flag.
legacy_pages.txt: @routes removed. TestRoutesFollowNavigation is enforced
from here on: every path of plan section Routes is registered with its
method, no pre-2.0 path is answered, and no template links to a path the
router does not serve.
Not in this step, left for the Overview page of stage 2: the outbound and
inbound domain summaries the mockup shows under the cards.
Tests: the six cards from one reading of the checks (values, levels, links,
and the unverified case); the two modes of the page - Overview without the
tables and actions, Health with them, each polling its own fragment.
go test ./... green except the two Windows-only failures in internal/domain
and internal/logtail.
Roadmap 15/27.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Stage 1, step 4 (plan § Routes).
Service (internal/app)
- Settings{Mode, Addresses, AuthIPRestrict, AuthAllowedIPs, Limit} is what
the form holds besides the login. CreateWithSettings creates the
application with all of it and removes it again if a later write fails;
SaveSettings validates the sender rule first and, if a later write
fails, puts back what it had already changed. The form produces a whole
application or leaves things as they were.
- Limit.Mode "" is "use the domain limit"; auto takes its ceiling from a
recalculation the caller supplies.
Routes and handlers
- GET, POST /outbound/domains/{id}/applications/new
GET, POST /outbound/domains/{id}/applications/{aid}
POST .../{aid}/password, .../{aid}/ratelimit/recalc, .../{aid}/delete
replace POST /domains/{id}/applications and the six /applications/{aid}/*
routes. lookupDomainApplication answers 404 when the application is not
in the domain named by the path - also when the caller may open that
domain.
- A generated password is the response to its POST only (no GET path) and
is sent with Cache-Control: no-store.
- The login is not renamed by the form; a changed login is a 400.
Old domain page: the three per-application forms (mode, IP restriction,
limit) are one form with one Save; "Remove limit" became the "Use the domain
limit" choice. The e2e client follows (edited, not run here).
Tests: service - create and save in every limit mode, and all-or-nothing on
a refused sender rule, a failed recalculation and a failed sender-map
rebuild; handlers - the form's parsing, one POST creating sender + IPs +
limit, four kinds of refused form creating nothing, save, the no-store
header on both password pages, and five application routes against another
tenant's application addressed through its own domain and through the
caller's.
With @routes removed by hand TestRoutesFollowNavigation now reports only
GET /server/health (step 5).
go test ./... green except the two Windows-only failures in internal/domain
and internal/logtail.
Roadmap 14/27.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Stage 1, step 3 (plan § Routes). BREAKING: every panel URL changes and the
old paths are removed without redirects.
Routes (web.go)
- /status -> /overview; /domains -> /outbound/domains with the domain's
settings actions under .../settings/ (reports, ratelimit, export);
/deliveries -> /outbound/log; /mail-queue -> /outbound/queue; /dmarc ->
/outbound/dmarc; /inbound -> /inbound/domains; /system-log ->
/server/log; /backup -> /server/backup (import under it); /users ->
/server/users; recheck and reload under /server/health/.
- One fragment name: <page>/fragment everywhere.
- Everything under /server/ is registered through server(), which wraps
the handler in globalOnly: the subtree is guarded once, where a new
route cannot forget it. Handlers keep their own requireGlobal.
- GET / leads to /overview, /outbound/domains or /inbound/domains.
- The /account -> /settings redirect is gone with /settings.
Account and Server > Settings
- handlers_account.go: GET /account and three POSTs - profile (username,
e-mail; no password asked, as designed), password (current password,
rate-limited like a sign-in, signs out other sessions), dmarc (the
user's default: hosted / account / custom / none, each validated).
This replaces step 1's bridge through the old Settings form.
- handlers_settings.go is the instance's page only: log retention and the
read-only rate limits, global role.
- One legacy template serves both pages by a flag until stage 2 gives each
its own: a new legacy page cannot be added to the ratchet.
Still to come inside stage 1, and why @routes stays on the ratchet:
- step 4: the application routes (one POST, under the domain);
- step 5: GET /server/health.
With @routes removed by hand TestRoutesFollowNavigation reports exactly
those and nothing else - no template links to a path the router lacks.
Old templates, tests and the e2e client are retargeted. The e2e module was
edited but not run here (no Docker, and its dependency could not be
downloaded on this machine).
Tests: Account (page, profile, password, DMARC default in every mode, a
domain administrator's own row only), Server settings (saved without a
password, range, global only), the /server/ subtree answered 404 by the
router for a domain administrator with All on both lists and for no
principal.
go test ./... green except the two Windows-only failures in internal/domain
and internal/logtail.
Roadmap 13/27.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Stage 1, step 2 (plan § Who sees what).
Store
- Reach{AllDomains, DomainIDs, AllInbound, InboundDomainIDs} is what a
domain user is assigned. CreateUser takes it; SetUserReach / ClearUserReach
replace SetUserDomains / ClearUserDomains and write both directions in one
transaction. A domain user with nothing on either list is refused
(ErrEmptyReach). With All set the rows of that direction are not stored,
so clearing All later brings nothing stale back.
- ListDomainsForUser honours all_domains; ListInboundDomainsForUser is its
inbound twin. Both filter in the query.
Authorization
- auth.Principal carries the whole reach: CanAccessDomain (now also true
with All), CanAccessInboundDomain, HasOutbound, HasInbound. An id or a
name on one list opens nothing on the other.
- Inbound handlers in three strengths: requireInbound (the list, filtered),
requireInboundDomain (one domain: page, DNS re-check, upstream,
recipients), requireInboundGlobal (add, delete confirm, delete). All
answer the same 404.
- domainIDSet and the DMARC report pages go through the principal instead
of looking at p.Domains themselves.
Panel (old templates, until stage 2)
- User form: two lists with All; inbound is offered only where the feature
is on, and saving the form with it off keeps the user's inbound reach.
Promotion to global drops the reach.
- Menu: a group with nothing assigned is absent; / sends an inbound-only
administrator to Inbound. The inbound pages hide add / delete from a
domain administrator.
Tests: every route of one inbound domain against five kinds of refused
request (another tenant's id, a missing id, outbound-only reach, an outbound
domain with the same id, one with the same name, no principal) with the 404
bodies compared; add/delete as every non-global user; the filtered list
including a domain added later; thirteen outbound routes against another
tenant's domain and application id; the user form; the reach in the store;
the principal matrix; the menu; the home redirect. Removing the access check
or the query filter makes the tests fail (checked by hand).
go test ./... green except the two Windows-only failures in internal/domain
and internal/logtail that predate this work.
Roadmap 12/27.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Stage 1, step 1. BREAKING: a 1.x database is not migrated.
Schema
- internal/store/migrations/ holds one file, 0001_init.sql, equal to the
2.0 schema; 0002..0010 are deleted (they stay in history at 1.9.5). No
admin table created and dropped, no ALTERs.
- users.email replaces users.dmarc_report_email: the address is the
user's own. The default DMARC report address is a user's setting
(dmarc_default_mode hosted/account/custom/none, dmarc_default_address);
the instance setting that mirrored the global user's field is gone.
- domains.dmarc_rua_user_id: a domain follows a named user's default, or
has an address of its own in dmarc_rua ('' = none). ON DELETE SET NULL
leaves it at "no reports", never at someone else's default.
- user_inbound_domains, users.all_domains, users.all_inbound_domains: in
the schema for the inbound delegation of the next step.
- users.role is 'global' / 'domain' (was 'domain_admin');
rate_limits.allowed_ips, unused since 1.9.0, is dropped.
Opening a database
- The baseline sets PRAGMA application_id ("SP20"). store.Open refuses a
file that has a schema version without it (ErrForeignSchema) and one
newer than the build. A 1.x file at user_version 10 would otherwise read
as already migrated. The refused file is left untouched.
Code
- Store.DomainDMARCRua is the one place a domain's report address is
resolved; dnscheck.ResolveDMARCRua and Store.GlobalDMARCReportEmail are
gone. dmarc.AllowedRecipients, the domain page, the dashboard and the
DNS re-check go through it.
- The old templates keep working on the new model until their pages are
restyled: the Settings form's address field is the account e-mail and
switches the user's default between it and none; a domain's "Same as
Settings" makes it follow the user who chose it; a new or imported
domain follows its creator.
- Domain export/import: a domain that follows a user exports no address.
Tests: the baseline pinned table by table (TestBaselineSchema), refusal of a
1.x and of a newer database, constraints and cascades of the new columns and
table, resolution in every default mode, the deleted-user case, and the three
bridged handlers.
docs/schema-migrations.md is rewritten for the baseline; roadmap 11/27,
schema-squash marked built.
go test ./... (Windows)
ok for every package except two that fail the same way before this commit:
internal/domain TestWriteLoadPrivateKeyRoundtrip (file mode on NTFS)
internal/logtail TestFollowTailsAndRotates (rename of an open file)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Follows the design: commit before it. panel.css gains the same four rules
theme.css did (TestPanelCSSIsTheMockupStylesheet holds them equal): the form
stretches the button across the user menu and the button is handed Bulma's
hover / focus / active variables. layout.html already had the form; only its
comment changes.
Checked in the browser on the kit page at 1440: Sign out is as wide as
Account and Help (147 px, 36 px tall), shows the pointer cursor, and is
tinted under the pointer like they are.
go test ./internal/web/...
ok github.com/mixeme/selfpost/internal/web
ok github.com/mixeme/selfpost/internal/web/auth
ok github.com/mixeme/selfpost/internal/web/handlers
ok github.com/mixeme/selfpost/internal/web/validate
ok github.com/mixeme/selfpost/internal/web/view
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Owner's instruction (2026-10-09): in the kit Sign out had no hover tint and
stood out among Account and Help. The cause was in the design, not in the
implementation: the mockups drew Sign out as a link, the panel signs out
with a POST, and Bulma tints only a.navbar-item.
- build.py: Sign out is a form with a submit button carrying navbar-item,
as the real shell has to write it; all pages rebuilt.
- theme.css: the form stretches the button across the menu and the button
is handed Bulma's own hover / focus / active variables, so it answers
like the entries above it. No width, no colour of its own.
check.py: 32 screens, 48 sp- classes, 0 violation(s); outlines.json unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Stage 0, steps 3-5, in one commit: the three were built together (the kit
page is what exercises the partials and the shell) and do not split into
states that build on their own. Written by Sonnet, reviewed by Opus.
- static/panel.css: shared/brand.css + panel/theme.css, rule for rule.
wordmark.svg and perf-edge.svg beside it; the stamp is logo.svg.
- templates/components.html: every partial of the plan's Three layers,
with its typed input in components.go. Wrappers are open/close pairs
(box_open/box_close, body_open/body_close, record_open/record_close).
Rich text is a Text value built from escaped runs - no field is
trusted HTML. status_tag and wbr_at are template functions; copy_field
is a partial whose button panel.js wires through data-copy.
- templates/layout.html: the shell as build.py draws it, the menu paths
written out literally, visibility from a typed Shell (role, feature
flags, the user's reach); the signed-out screen as a second define.
Not in the 2.0 menu: Inbound Log / Queue / Quarantine / Filter lists
and Preflight.
- GET /server/components (global role, 404 otherwise): the mockup
components page rebuilt from fixtures (kit.go) by the real templates.
- The old panel steps aside unchanged: panel.css -> legacy.css,
layout.html -> layout_legacy.html; a page uses the new layout iff it
is off legacy_pages.txt.
- legacy_pages.txt: @kit removed - the kit is under the guards for good.
Differences from the mockup components page, for the owner:
- the head button is Help, not "Cover sheet"; two "Example" links are
dropped (no fixed URL), two point at /outbound/domains and /login;
- the page sits in the Server group, so the sibling strip shows;
- Sign out is a POST form whose button carries navbar-item: same size
and type as the links, but without Bulma's hover tint;
- a head's actions are always wrapped in div.buttons.
Evidence
- docs/assets/panel-redesign/evidence/components.png: 1440 px wide, full
height, headless Edge. Taken from the real templates served locally,
NOT from the container (no Docker on this machine); the page reads
only its own fixtures and embedded assets.
- panel/audit.js on the kit page, 1440 and 1280:
{"pages":1,"widths":[1440,1280],"textsChecked":516,"problems":[]}
- go test ./internal/web/... (with @kit removed)
ok github.com/mixeme/selfpost/internal/web
ok github.com/mixeme/selfpost/internal/web/auth
ok github.com/mixeme/selfpost/internal/web/handlers
ok github.com/mixeme/selfpost/internal/web/validate
ok github.com/mixeme/selfpost/internal/web/view
TestRoutesFollowNavigation stays skipped behind @routes (stage 1).
Roadmap 10/27.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Stage 0, step 2. The guards land before any template is restyled.
Guard tests (guard_*_test.go under internal/web) read their rules from the
accepted mockups instead of copying them - the Bulma subset and WIDTH_OK
from check.py, the shell's classes from build.py, the stylesheet from
theme.css + brand.css, the skeletons from outlines.json:
- TestPanelClassVocabulary, TestTemplatesUseComponents,
TestPanelPageStructure, TestPanelOutlinesMatchMockups,
TestRedesignedPagesLoadOnlyTheKit: hold every page that has left
legacy_pages.txt to its mockup, in source and as rendered.
- TestPanelCSSIsTheMockupStylesheet, TestPanelCSSContract: panel.css is
brand.css + theme.css rule for rule.
- TestOutlinePortMatchesCheckPy: the Go port of outline() agrees with
outlines.json on all 33 screens.
- TestLegacyRatchetMatchesTheEngine, TestVendoredAssetsArePinned,
TestStaticHoldsOnlyTheKnownAssets.
- TestRoutesFollowNavigation: the 2.0 route table, no old path answered,
no template link the router does not serve.
The kit tests are skipped while legacy_pages.txt lists @kit and the route
test while it lists @routes; both were run with the entry removed and fail
where expected (no kit yet, old route table).
.github/scripts/design-first.sh, run per commit in test.yml: a commit that
touches the mockups or a guard touches nothing else under internal/web;
legacy_pages.txt only shrinks; a changed template is not, whitespace aside,
any template from history. Exercised on a scratch clone for each case.
Plan: Enforcement and Against resurrected pages describe the guards as
built (the ratchet is a file beside the guards, not a list inside them).
Roadmap 7/27.
go test ./internal/web/...
ok github.com/mixeme/selfpost/internal/web
ok github.com/mixeme/selfpost/internal/web/auth
ok github.com/mixeme/selfpost/internal/web/handlers
ok github.com/mixeme/selfpost/internal/web/validate
ok github.com/mixeme/selfpost/internal/web/view
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Stage 0, step 1. Nothing links the new files yet; the old panel is unchanged.
- static/bulma.min.css: css/bulma.min.css of npm bulma@1.0.4, byte for
byte (tarball integrity checked against the registry;
sha256 67fa26df1ca9e95d8f2adc7c04fa1b15fa3d24257470ebc10cc68b9aab914bee).
- static/tabler-icons.css + tabler-icons.woff2: @tabler/icons-webfont
3.49.0 subset with fontTools pyftsubset to the 57 icons the mockups use
(the plan said 58; the count is 57).
- NOTICE and development.md § Vendored front-end list both, MIT.
- Roadmap: panel-redesign is agreed (owner, 2026-10-09), 6/27; the entry no
longer says 29 screens or that Health is undecided.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
Groundwork the guard tests of stage 0 stand on; no behaviour changes.
- internal/web/view/legacy_pages.txt: the redesign ratchet. Every page of
today's panel is listed as still old, plus @kit and @routes for the two
pieces of work whose tests stay quiet until they land. The list only
shrinks (enforced by CI in the next commits).
- internal/web/view/fixtures_test.go: the registry of fixtures the guards
render redesigned pages with. Empty until the kit page exists.
- web.go: route registration moves out of Handler() into muxes(), so a
test can ask the real muxes which pattern answers a path.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
CI-only cut. The release workflow deleted the X.Y.Z-amd64 / X.Y.Z-arm64
package versions after the manifest merge; those versions are the images the
X.Y.Z manifest points at, so pulling 1.3.0 failed with a missing manifest.
No product code changes.
Fixed:
- release.yml: the per-arch tags stay in GHCR. GHCR cannot drop a tag
without deleting the image behind it.
- release.yml: the merge job now removes only untagged package versions
that no tagged manifest references (leftovers of publishing a version
again); any failed lookup aborts the step before anything is deleted.
- development.md release description and version table, image pin 1.9.5
in compose, README, guide.
Full detail: CHANGELOG.md (1.9.5).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Plan-only cut, owner-requested. The redesign goes first as 2.0.0 and
inbound-antispam-panel follows as 2.1.0 on the 2.0 baseline (migration
0002_inbound_spam_log.sql), built from the redesign's component kit so its
two screens are styled once. No product code changes.
Changed:
- panel-redesign plan: Component kit section and stage 0 (partials with
typed inputs in components.go, one panel.css, /server/components kit
page, acceptance before any page is restyled); staged checklist (5/27);
Health decided inside 2.0, queue pages left as old content in the new
shell.
- panel-redesign plan: Against resurrected pages - shrink-only
legacyPages ratchet in the guard tests enforced by CI, the outline
golden, the old-path test and a history-blob check, so an old template
from git history cannot pass as a redesigned page.
- roadmap: new candidate delivery-log-storage - the Delivery log box on a
message page is empty for anything older than the last daily rotation
because the lines are grepped from the current mail.log on each view
and never stored.
- development.md version table, image pin 1.9.4 in compose, README, guide.
Full detail: CHANGELOG.md (1.9.4).
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Panel-redesign prep, docs only: stage 1-2 mockups on all 33 screens, the accepted header/type-scale/contrast corrections, the account e-mail and inbound-delegation decisions, and the routes/quarantine/preflight roadmap screens. Nothing in `internal/web` changes; the design stays `candidate` (3/21) until the redesign itself ships as `2.0.0`.
Added:
- Panel redesign, account e-mail and notifications: a user's e-mail is its own
- Panel redesign, contrast: a sweep of every text node against its real ground
- Panel redesign, header (owner's choice: variant A, with the perforation on a
- Panel redesign, type scale: about 40 % of the panel's text measured under
- Panel redesign, header comparison sheet
… (8 more in CHANGELOG)
Full detail: CHANGELOG.md (1.9.3).
Review pass over `1.4.0`…`1.9.1`: port 25 no longer shares the submission milter chain, DMARC reports are validated before they are stored, and the auto rate limit is correct for a non-default level-1 window. Schema migration `0010` (index only). Upgrading is a tag bump.
Added:
- SQLite migration `0010_send_log_app_login_index.sql` —
- DMARC aggregate reports are accepted as **zip** as well as gzip or plain XML
- e2e: `checkPort25ConfigCombinations` regenerates the Postfix configuration
Changed:
- README hero line and a **Positioning** block in `docs/product.md` record the
Fixed:
- The setup account's password is hashed at bcrypt cost `12` like every other
- Login/setup rate limiter: protecting blocked buckets from eviction is now
- `postqueue -p` output is bounded to 4 MiB (the timeout alone left the size of
- [schema-migrations.md](docs/schema-migrations.md): the schema section was
- postqueue reader is now time-bounded: `postqueue -p` runs under a 5-second
… (11 more in CHANGELOG)
Changed:
- [roadmap.md](docs/roadmap.md) now carries everything the 2026-08-19 code
- Extracted `internal/configsafe` — the empty-value and forbidden-character
- Build, deploy and migration comments no longer cite the archived
- Dropped the `parseDomainRateLimitForm` / `parseAppRateLimitForm` pass-through
- Extracted shared helpers: `internal/atomicfile` (atomic temp-write+rename) and
… (7 more in CHANGELOG)
Removed:
- Unused legacy asset `docs/assets/icon.png` (superseded by the four brand SVGs
- Plan documents for already-shipped features: `domain-stats-auto-ratelimit`,
Full detail: CHANGELOG.md (1.9.2).
Port-25 Postfix startup when inbound relay, DMARC ingest, or the inbound antispam milter are enabled. Schema-migration reference and roadmap plans for inbound antispam and quarantine. Upgrading from 1.9.0 is a tag bump; no migration.
Added:
- [docs/plans/inbound-antispam-panel.md](docs/plans/inbound-antispam-panel.md) —
- [docs/plans/inbound-quarantine.md](docs/plans/inbound-quarantine.md) — candidate
- [docs/schema-migrations.md](docs/schema-migrations.md) — living reference for the
Changed:
- [docs/roadmap.md](docs/roadmap.md) — `inbound-antispam-panel` (agreed, 1/12)
- [docs/development.md](docs/development.md) — plan checklists may override
- [docs/development.md](docs/development.md), [docs/roadmap.md](docs/roadmap.md),
Fixed:
- `build/postfix-config.sh`: enabling inbound relay, DMARC ingest, and/or the
Full detail: CHANGELOG.md (1.9.1).
Application sending controls: client IP allow-list for authorization, and level-2 rate limits that override the domain ceiling per application.
Added:
- Per-application **client IP allow-list** on the domain page: when enabled,
- SQLite migration `0009_application_auth_ips.sql` (moves legacy trusted IPs
Changed:
- Application level-2 rate limit **overrides** the domain limit for that login
- Domain export/import carries `auth_ip_restrict` / `auth_allowed_ips`; legacy
- Panel help, Settings, and [guide.md](docs/guide.md) describe override semantics
- [guide.md](docs/guide.md): full-backup section documents all four capture
Full detail: CHANGELOG.md (1.9.0).
Built-in operator documentation in the panel: a CSS-only help drawer and Help page so Status checks and domain controls are explained without leaving the panel.
Added:
- **Help** navigation entry and `/help` index page.
- CSS-checkbox help drawer (no extra script): seeded topics for Status (Machine,
- «?» entry points on Status and domain card heads open the matching drawer
Changed:
- [guide.md](docs/guide.md) documents the boundary: in-panel help for card
Full detail: CHANGELOG.md (1.8.0).
Optional DMARC aggregate report ingest: accept `rua=` mail on port 25, parse gzip/XML into SQLite, and show summaries in the panel.
Added:
- Optional DMARC aggregate report ingest (`DMARC_REPORTS_ENABLE`): port 25
- Per-domain **SelfPost hosted** `rua=` mode when ingest is enabled
Full detail: CHANGELOG.md (1.7.0).
30-day sending statistics per domain and application, plus optional auto level-2 rate limits derived from average send rate.
Added:
- Domain page **Sending statistics** card: total messages, peak and average
- Level-2 rate limit **Manual** / **Auto** mode on domain and application
- Domain export JSON now includes rate limits (`mode`, `auto_multiplier`,
Changed:
- Auto mode uses the level-1 window (`RATE_LIMIT_WINDOW_SECONDS`); manual
Full detail: CHANGELOG.md (1.6.0).
Optional inbound relay (backup-MX / forwarder) on port 25, off by default. Mail is accepted only for domains configured in the panel and forwarded to an upstream; there are still no mailboxes. A dedicated security review of the inbound path is still open on the plan (Fable).
Added:
- `INBOUND_RELAY_ENABLE` (default `false`): when true, Postfix listens on
- Panel **Inbound** pages (global administrator): domains, upstream host/port
- Optional `INBOUND_ANTISPAM_MILTER` on the inbound listener only, plus
- Inbound rate limit (`INBOUND_RATE_LIMIT_MESSAGES_PER_IP`, default 20) and
Changed:
- Compose publishes host port 25 even when the flag is off (nothing listens
- Full-backup restore Resync also rebuilds inbound maps when the flag is on.
Full detail: CHANGELOG.md (1.4.0).
[1.3.0]
Security and quality after 1.2.5: domain-admin send-log authorization, fail-closed sign-in and application delete, level-2 rate-limit race fix, restore Resync, expanded tests, operator docs, release CI, and OFL for IBM Plex. Upgrading from 1.2.x is a tag bump; no migration.
Added:
- licence: the SIL Open Font License 1.1 text now travels with the IBM Plex
- docs: agreed roadmap item **queue-retries** — show this Postfix's retry
Security:
- The independent security review of the send-log authorization and
Fixed:
- test (e2e): send-log status scrapers follow the badge markup in
- security (panel): the Deliveries list is scoped to a domain administrator's
- mail (level-2 rate limit): the ceiling is no longer overshot by messages that
- panel (sign-in): a session that cannot be written to the database no longer
- panel (applications): deleting an application removes its SASL credentials
… (2 more in CHANGELOG)
Changed:
- docs: operator and as-built docs aligned with the code after a full
- docs: [guide.md](docs/guide.md) reorganised into **Installation**, **Instance
- docs: the 2026-08-13 full-tree review plan is complete — every phase (P0–P7)
- licence: [NOTICE](NOTICE) tells modifiers to update `SourceURL` in
- ci: the release image is published only for a **published** GitHub Release
… (9 more in CHANGELOG)
[1.3.1]
Retry policy in the panel, persistent Postfix queue, and self-contained full backups. Mail queue and delivery history show this Postfix's first retry, backoff cap, and queue lifetime. Deferred mail survives container recreate. Full backups now archive `data/`, compose, `.env`, and `certs/`; the archive layout breaks 1.3.0 flat backups (unpack those with `tar xzf backup.tar.gz -C ./data` as before). Upgrading from 1.3.0 is a tag bump; no schema migration.
Added:
- Mail queue and a delivery's history show this Postfix's retry policy
- docs: **Plan checklists** in [development.md](docs/development.md) — format,
- docs: **Implementation checklists** with model routing in
- docs: roadmap candidates **send-log-retention** (panel Settings for delivery
- Postfix queue under `/data/postfix/queue` — deferred and active mail survive
… (1 more in CHANGELOG)
Changed:
- **Breaking:** full-backup archive layout — paths are prefixed with `data/`;
- docs: operator guide, architecture, security, and backup UI updated for
- docs: HTML mockups for a full panel UI refresh
Changed:
- docs: panel UI mockups after review — Status restored to live readings
- docs: panel UI design system and mockups rebuilt as separate pages
- docs: panel UI mockups — add-domain sits in the list card (Domains and
Fixed:
- ci (GHCR): per-arch package tags (`X.Y.Z-amd64`, `X.Y.Z-arm64`) are dropped
Squashed CHANGELOG sections: [1.3.0], [1.3.1].
Full detail: CHANGELOG.md (1.3.0, 1.3.1).
[1.2.0]
The second MINOR after 1.0.0: domain administrators with per-domain scope, a panel visual refresh on the SelfPost palette, and refinements to navigation and the send log. Upgrading runs one SQLite migration (`0005_panel_users`); the single administrator becomes a global user.
Added:
- panel: **domain-admin role** — global administrators manage panel users and
Fixed:
- Docker build: `LICENSE` is no longer excluded by `.dockerignore`, so the
- panel: on signed-in pages with a narrow card (**Settings**, the user form)
Changed:
- panel: navigation session icons are distinct — **Settings** uses a gear,
- panel: `/` redirects domain administrators to `/domains`; global users still
- panel: **Sign out** in the navigation column uses the same type size and
- panel: **visual style** brought in line with the SelfPost mark — brick accent
- panel: the send log's **status is a badge**, in the same ok/warn/error/unknown
… (1 more in CHANGELOG)
[1.2.1]
Panel refinements after 1.2.0: navigation icons, status-badge centreing, drill-down back-link placement, and safeguards for the sole global administrator. Upgrading is a tag bump; no migration.
Fixed:
- panel: status badge text sat low in the box (and below the heading or label
- panel: the **Users** navigation icon was two full silhouettes with staggered
- panel: the **Settings** navigation icon was a sun-with-rays (circle plus
- panel: the user create/edit form placed «Back to users» at the bottom of the
Changed:
- panel: the user edit form disables role change and delete for the only global
- panel: the user create/edit form hides **Assigned domains** when the role is
- panel: **Settings** shows panel credentials and DMARC aggregate reports side
[1.2.2]
Status page layout after 1.2.1: paired cards in a wide column, denser machine and check copy, and a user-form checkbox fix. Upgrading is a tag bump; no migration.
Changed:
- panel: **Status** is wide again so paired cards fill the column. Layout:
Fixed:
- panel: cards inside `.split` used `margin: 0 auto`, which in a CSS grid
- panel: on the user create/edit form, **Assigned domains** checkboxes stacked
[1.2.3]
Domain detail layout and panel polish after 1.2.2. Upgrading is a tag bump; no migration.
Changed:
- panel: the domain detail page is wide with paired cards (DKIM+SPF ‖ DMARC;
- panel: Domains list — **Add domain** sits beside the domain field; the
- panel: page URLs, browser titles, and headings are aligned — **Settings** is
Fixed:
- image: `mail.log` rotation no longer silently stops when the build context
[1.2.4]
Level-2 rate-limit semantics inverted after 1.2.3, plus a small DNS field height fix. Upgrading is a tag bump; no migration.
Changed:
- rate limiting (level 2): domain ceilings apply to every client IP (no IP
Fixed:
- panel: on the domain DNS status grid, the Type (TXT) field height matches
[1.2.5]
Rate-limit form polish after 1.2.4. Upgrading is a tag bump; no migration.
Changed:
- panel: rate-limit forms refined — level-1 backstop as muted copy on the
Squashed CHANGELOG sections: [1.2.0], [1.2.1], [1.2.2], [1.2.3], [1.2.4], [1.2.5].
Full detail: CHANGELOG.md (1.2.0, 1.2.1, 1.2.2, 1.2.3, 1.2.4, 1.2.5).
The first MINOR after 1.0.0: send-only DMARC guidance in the panel, AGPL packaging on every page, and an internal split of `internal/web` ahead of domain-admin work. Upgrading runs one SQLite migration (empty defaults; existing DNS guidance is unchanged until you set a report address).
Changed:
- `internal/web` split into subpackages (`web/view`, `web/auth`, `web/validate`,
Added:
- panel: DMARC guidance for send-only relays — the suggested `_dmarc` record
- AGPL packaging hygiene: [NOTICE](NOTICE) names the copyright holder and the
- [docs/roadmap.md](docs/roadmap.md) — candidate item **visual-style** (panel
Full detail: CHANGELOG.md (1.1.0).
[1.0.0]
Added:
- A **Delivery log** on each delivery's page (`/deliveries/{id}`): the
- A **History** block on the same page: the journal's two timestamps stated as
Fixed:
- Postfix config: copy TLS cert/key from the (often `:ro`, host-owned) mount
- Postfix config: allow `maillog_file` under `/data` via
- E2e: read `/data/setup-token` via `docker compose exec` (file is `0600`
- Entrypoint: check `SELFPOST_HOSTNAME` before `/data` setup (so a bad identity
- E2e gate: wait for host-published `/healthz` before panel setup, and stop
… (3 more in CHANGELOG)
Changed:
- Document what `.spbk` and `.spde` stand for (SelfPost backup / SelfPost domain
- Docs aligned with the code: setup URL is `/setup/<token>` (README and
- Full backups no longer carry `/data/log`. It is Postfix's raw log plus its
- Monitoring screens (status, mail queue, system log, deliveries) use adaptive
- Documentation package consolidated into `docs/development.md`: Documentation
… (4 more in CHANGELOG)
[1.0.1]
A documentation and packaging release: no change to the mail path, the database, or the on-disk layout. Upgrading is a tag bump.
Added:
- `SECURITY.md` — how to report a vulnerability privately (GitHub private
- [docs/plans/](docs/plans/) — one document per agreed extension: the optional
Changed:
- The panel's **Account** entry is now called **Settings** — nav link, page
- The signed-in name in the panel's nav is now labelled `User:`, so it reads as
- [docs/product.md](docs/product.md) reframes the future line: the inbound
- [docs/security.md](docs/security.md) is now in English, matching the rest of
- The README documentation table now points at `SECURITY.md` for reporting a
… (2 more in CHANGELOG)
Fixed:
- The panel's static assets are served with a content ETag and
Squashed CHANGELOG sections: [1.0.0], [1.0.1].
Full detail: CHANGELOG.md (1.0.0, 1.0.1).
A page per delivery (`/deliveries/{id}`), reached from the *Details* link on
Added:
- A page per delivery (`/deliveries/{id}`), reached from the *Details* link on
- A **DNS** badge in the domain list, one per row, carrying the same
- Machine metrics on the status page: a **Machine** card reporting the
- An index of the current page's own sections in the navigation column, for the
Changed:
- Every panel page is laid out in one column of the same width, so moving
- The mark's small-size variant — the tab icon and the `SP` initials it
- The stamp's `SELF-HOSTED SMTP RELAY` line is set at 11.5/0.15 instead of
- The delivery log lists what identifies a message and nothing else: time,
- Subjects are decoded for display as well as on the way in, so the rows the
… (4 more in CHANGELOG)
Full detail: CHANGELOG.md (0.6.0).
mixandClaude Opus 5.5