The card's, note's or step's own id goes with a suggestion, and a sealed
file goes along readable only where it is under the key of that item's
tree: a file of another private tree, named by a row the instance
planted, is left out and never uploaded readable (the short reading of
22 Oct, finding 12).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
Run on a1adc18, every test of a fixed finding failed but two: the replayed
sealed row now comes back under its own id, as the instance would replay
it (a copy under another id never opened), and the pending make-readable
now watches the files, as the old code's sweep sealed its rows again
after opening them. The older readable record against a newer private one
passes there too: it guards what was right already.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
Publishing is refused on the device where anything in the tree is sealed,
before any tag's name goes out: the tree and the tags' links are the
instance's to change. A suggestion leaves sealed files out, as the
instance could name another private tree's file through a planted row.
Docs: ADR 0019's "The short reading (22 Oct 2026)", and the general
residual risk that any device can be held at an earlier genuine record;
Learn's AGENTS.md.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
A rewrite or sweep changes only rows under keys whose signed ID names the
studyset rewritten (or none), so making a studyset readable never opens
another tree's rows the instance moved into it, and a sweep never seals
them under its own key; the person's own moves alone re-key rows of other
keys. An item's key is never replaced by an ended key, nor by another
studyset's while the one kept holds, and is written with before the
tree's. A pending rewrite the latest record no longer asks for is
dropped. A file that won't move is counted and owed, the tree partly
readable meanwhile, and each step of the upkeep fails alone. Settings
carry a version vector (each device's count of writes): a row is refused
as a rollback only when it brings news from no device. Sweeps look rows
up by their key and skip those known unmovable; a sync is asked for only
when something was written. Rewrites run one at a time and write back
only rows unchanged since they were read. A record is made later than
the latest known, whatever the clock. An unreadable ledger is kept aside
and nothing private is written until the keys are opened again; whether
the account has keys is the register's to say.
Tests (each the reading's probe): a moved card under a studyset made
readable, a sweep after a move, a replayed sealed row under an ended key,
a pending make-readable over a later private record, a file that won't
move, two devices' settings at once, an older readable record on a fresh
device, a clock behind; the withheld-record test withholds the tree's
record alone.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
The reading of 22 Oct, on the device's side. Private cards never move
into a readable studyset: the card browser asks to make it private first.
Locked studysets are left out of the move sheets and the destinations
words are added to. A studyset whose key isn't on this device says to
sign in with the password here, not that it is lost, and can still be
archived; creating one there says why it can't. The marker shows the lock
for a private studyset published before, and a shield with a warning
while something in it is still readable or unsealed, with Finish in its
dialog. The library's offer covers every top-level studyset, Recently
deleted and orphaned ones too, says what couldn't be sealed, and its
answer follows the person in their settings (schema 11).
The full-window editors carry the studyset they were opened from, and an
account with keys adds no file where none is known (a new studyset's
description takes pictures once it is made). A file no live row names is
never uploaded, and deleting a row clears what it owed. Sealed rows are
indexed by key and hash. A suggestion from a private copy uploads
readable copies of its files for the author; accepting swaps whole
`asset:` values only.
Docs: ADR 0019's "The reading (22 Oct 2026)" with the decisions, what
changed and the residual risks; "As built" and Learn's AGENTS.md brought
up to date.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
The reading of 22 Oct, against an instance that alters data. A data key's
ID now says what it is for (a studyset's key naming it, the person's key,
or a record that a studyset was made readable), and the ID is sealed with
the key and signed by the account, so the instance can't relabel a key or
make one up; the row's own columns are no longer trusted. A studyset is
private while the latest signed record about it says so, and the device
keeps the latest it saw in a ledger in the person's folder (privacy.json),
so a record withheld or a row replayed changes nothing; a tree's key is
the nearest studyset's with such a record, itself or above. An item keeps
the key it was last seen sealed with unless this device moves it or a
signed record makes its tree readable.
A readable row a device holding the keys writes carries the account's
signature in place of a nonce; one nobody signed is locked wherever
privacy holds (in private trees, for items once sealed, and for tags and
settings once sealed), so the instance can't plant content, confirm a
guessed tag name by folding, or switch Count me on. Signed readable rows
from a device that hadn't heard are shown and sealed anew.
Rewrites are kept as under way until done, write the top studyset last,
hold writes into their tree meanwhile, and are finished as the data opens
and after each merge; a sweep seals what a stale device wrote. Cards moved
into a private tree are sealed with their files first; cards of a private
tree are never moved into a readable one. A private topic moved out of
its tree keeps its privacy with a key of its own. Settings carry a counter
inside their sealed payload, and a lower one than seen is refused. The
person's readable tags and settings are sealed once. A device without the
keys of an account that has them writes nothing that would be private and
reports nothing for the learner count; an anonymous person's keys kept but
unreadable are never replaced. Publishing sends only the names of the
tags the release carries. Key refusals are remembered per row.
Tests on the real gateways over a local database: a replayed readable top
row, a withheld key record, a moved parent, a planted revision, a signed
stale revision, card and topic moves, an interrupted rewrite, a settings
rollback, a device without keys, planted tags.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
An upload said to be encrypted must start as a sealed file does (HLKF,
version 1), so a readable file can't skip the PDF check by saying so. A
person's room counts the files a live asset row names and those claimed
within the week of grace, so a private studyset's rewrite near the limit
no longer fails for a week while pruning keeps the replaced files. Copy
reports are limited to 60 an hour per person. Tests: what goes with the
person, `counted` refused to anonymous people.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
The developer's decision of 21 Oct, on the library's cards and the
studyset's Share in its action row alike:
- a lock: private, end-to-end encrypted;
- an open lock: readable by the instance, not published;
- a globe: published.
The lock and the open lock open one dialog saying what that means and
changing it (Make non-private; Make private, or Publish…, which opens the
Publish sheet, while it isn't private); a topic's dialog says its privacy
is its tree's. The globe opens the Publish sheet, whose status for a
studyset not published now says so ("Not published") with an open lock,
as "Private" means something else now. Texts in en and de.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
- `AssetRepository` seals a file of a private studyset with its tree's key
for its asset row before hashing it (`AssetSealing`, which `Privacy`
does): the row's hash and size and the file uploaded are the sealed
file's, the upload says it is encrypted, and the row's payload, sealed
too, carries the file name and the hash of what was sealed (asset
schema 2). A download is checked against the row's hash, opened for its
row and checked against what was sealed. This device keeps the sealed
file to upload and the opened one, by its own hash, for what reads files
by path.
- The same file is a row per key: in a private and a readable studyset, or
two private trees, it is two.
- Every place adding a file names its studyset: the editors' pictures and
sounds (`StudysetFiles`, which every screen of a studyset and the PDF
reader set), PDFs and screenshots, Anki imports, copies and their updates
(a copy's top studyset is made first, so its files take its key), and
accepted suggestions (by their item's tree).
- Making a tree private or not, or moving a topic into another tree, moves
its files: each under another key is added anew under the tree's, the
content names the new row (whole `asset:` values and a PDF's file, never
text), and the old row is blanked and deleted once nothing else names it.
- Tests: a private studyset's file never uploads readable; a sealed file
moved to another row, or under a key not open here, doesn't open; a
changed file is refused by its hash; one row per key.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
Learn's AGENTS.md says how private studysets work (keys, the envelope,
locked rows, rewriting a tree) and what changed for the connector,
publishing, uploads and the learner count; ADR 0019 records what was built
otherwise than decided or not yet (a private studyset's files); the
halkyon_app README describes the content cipher and device keys.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
New studysets are private; those from before stay readable until the
person chooses. The library offers once, on this device, where it can
encrypt and some top-level studyset isn't private: Make all private, or
Not now; either answer is remembered (`private_offer` in the person's
folder) and the offer goes. Each studyset's menu changes it later.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
- A top-level studyset's options make it private, with everything inside
it, or non-private again, after a dialog saying what changes and that
the instance's backups keep readable copies until they expire; a device
without the account's keys says how to get them there.
- A locked studyset (its key doesn't open here) is titled Locked, says
why when opened, and offers only Delete.
- Publishing names the person's tags as the device reads them, and says
what to do when the studyset is private or a tag has no name.
- The export sheet says an archive is readable by whoever has the file.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
Count me stays in the person's settings, which follow them to every device
and are now encrypted; after each sync the device mirrors it to the
instance (`learningCount.setCounted`) and, only while it is on, reports
which publications their copies come from, topics with them
(`reportCopies`). Taken back, the instance forgets the copies. Only people
with an identity report, as only they are counted.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
- `ContentKeys` opens the person's data keys (`content_key`) with their
account's keys on the device, an anonymous person's device keys made as
their data first opens; it makes a private studyset's key and the
personal key. Without the account's keys nothing opens and nothing new
is private (D8).
- `PayloadEnvelope` is the one place payloads are sealed and opened, each
for its table, row and column, so a payload moved to another row never
opens; one that doesn't open is `Locked`, never half-read.
- `Privacy` gives each gateway the key of the row's tree (its top-level
studyset's), and rewrites a tree under another key, or readable, in
place (`rekeyTree`).
- The studyset, card, note, learning path, path answer and PDF gateways
seal with their tree's key and open what they read: a new top-level
studyset is private where the device holds the keys, a new one inside a
tree takes its privacy, and one moved into another tree is rewritten
under that tree's key. Tags and settings are sealed with the personal
key; a tag or settings row that doesn't open here is left out, so
nothing folds or deletes it. A studyset whose key doesn't open shows as
locked (`Studyset.isLocked`); other rows that don't open read as blank.
- Every payload codec round-trips sealed, and refuses another row.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
- `content_key`, a synced table in the person's space, keeps their data keys
sealed to their account (a private studyset's, their personal one);
`learning_path_answer` gains `keyId` and `nonce` for answers in private
studysets. Additive only, on the server and the device.
- Publishing refuses a private studyset, and a tree holding a private row
(`PublishRefusal.private`); a private studyset elsewhere no longer breaks
publishing a readable one. Encrypted tags are named as the device names
them (`tagNames`): a name for a tag that isn't the person's, or an
encrypted tag of the release without one, refuses it
(`tagNamesMismatch`); a readable tag keeps its own name.
- The connector leaves private studysets out and says how many there are.
- The learner count reads no payload any more: Count me is kept by the
instance (`counted_learner`) and the device reports which publications
its copies come from (`learning_copy`, accepted only while counted, each
report replacing the last); only the reporter's own studysets count.
Deleting a person deletes both.
- An upload says whether its file is encrypted: only a readable one over
20 MB is checked to be a PDF, and an encrypted one may be 4 KiB larger.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
The sheet holds snack bars of its own (`hasSnackBars`), so a failed
deletion, sign-out or reading of the account says so on it rather than
behind its barrier.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
ADR 0017's rules as they stand, halkyon_app's README and Learn's
AGENTS.md: where an instance offers passwords, the apps offer no Add GitHub
and no hint on adding it, an anonymous person's Sign in with GitHub reaches
only the account someone has, and every refusal says why.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
Someone identified was offered Add GitHub wherever GitHub signed people in,
and on an instance offering passwords the browser round trip ended in
"Couldn't sign in with GitHub.", as the instance adds GitHub to nobody there
(ADR 0017, D2; the end-to-end test of 21 Oct). The profile now shows the
GitHub row there only to whom it signs in already, and an anonymous
person's Sign in with GitHub, which still reaches the account someone has,
says why it didn't: a GitHub account nobody has, with the password beneath
in its place, or one that signs in with its password now.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
Where an instance offers password sign-in, GitHub is added to nobody (ADR
0017, D2), yet the start screen still said "then add GitHub" after a GitHub
account nobody has, and every refusal but an unknown account read
"Couldn't sign in with GitHub." (the end-to-end test of 21 Oct).
- The gateway tells from one read of the basics whether GitHub signs in and
whether it is added to people (`gitHubOffer`, in place of
`gitHubClientId`); the GitHub bloc keeps it (`Ready.linking`) for the
profile, with why an attempt failed (`Ready.problem`).
- The instance's `usePassword`, `deprecated` and `ended` become problems of
their own (`gitHubSignInProblemOf`), worded by `gitHubProblemText` and on
the start screen; an anonymous person refused a link is told the GitHub
account is nobody's and offered the password instead.
- Where passwords are offered, the start screen offers an account to create
in place of the app's hint on adding GitHub.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
The text for an account that signs in with GitHub alone now names the
deletion 30 days after GitHub's end (f7a3251); Learn's test of its
profile looked for the words before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
ADR 0017 gets "The short reading": its findings, the developer's four
decisions of 21 Oct, and what changed. "The rules as they stand" says a
reset takes effect 14 days after it was asked at the latest, that a member
keeps the session a reset signed in and deletes nobody while the register
is out of reach, and no longer claims every secret is kept out of every
capture: Android holds them back, Windows from 10 2004, macOS from
screenshots but not ScreenCaptureKit on macOS 15 and later, iOS covers
the app switcher's picture alone, Linux nothing. The module's and
halkyon_app's READMEs and Learn's AGENTS.md follow.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
- Deleting an identified account whose security can't be read (the
instance out of reach) says so and deletes nothing; it went on to a
deletion by the session alone, refused for an account with a password,
and the person saw a spinner, then nothing. A refusal still leaves the
session's own deletion.
- InstanceChoice's sharesStorage is required, so no app leaves it out and
removes a folder another instance uses.
- A PasswordUnlock forgotten is spent: its secrets read again fail
(StateError) rather than make a proof or a wrapping of zeros.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
Before Windows 10 2004, SetWindowDisplayAffinity takes
WDA_EXCLUDEFROMCAPTURE without failing and acts as WDA_MONITOR, so the
fallback on a failure never ran and its comment was wrong (the short
reading). Both apps' runners now choose by the real version, RtlGetVersion
(GetVersionEx tells what the manifest claims): WDA_EXCLUDEFROMCAPTURE from
build 19041 on, WDA_MONITOR before. Checked for syntax only: nothing here
builds for Windows.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
The short reading's H1: open(2) takes its mode among variable arguments,
which Apple's Arm passes on the stack, so bound with three fixed ones it
read the mode from whatever the stack held: a phrase or recovery codes
saved on an Apple silicon Mac could be readable by everyone. It is bound
as variadic now (VarArgs), and every file written is made 0600 through its
descriptor (fchmod), checked.
M1: macOS's sandbox lets an app reach the file the person chose and
nothing beside it, so the file made beside it was refused and nothing
could be saved; Learn's sandbox allowed reading chosen files alone, so its
exports never saved on macOS either. There, writePrivately now writes the
chosen file itself (decided by the developer on 21 Oct): opened without
following a link or waiting on a pipe and without truncating, refused with
nothing written unless fstat finds a file of the own user's, then made
0600 and emptied. Both apps' entitlements grant reading and writing the
files the person chooses. Linux keeps the new file renamed over it.
The reading's nits: the buffer is wiped whatever happens, an interrupted
call is made again (EINTR), the file made beside a chosen one is named
apart from it so a name as long as a folder takes still saves, the folder
is flushed after the rename, and Linux reads a file's type and owner with
statx, alike on every architecture, where __lxstat's version differed
(riscv64's was wrong). The test of another user's file is skipped as root.
The in-place path is tested on Linux; none of it has run on a Mac.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
ADR 0017 gets "Reading the fixes (21 Oct 2026)": each finding and what
was done, and "The rules as they stand" brought up to date (resets take
effect once 72 hours outside hold-outs passed and lapse 30 days after
they were asked; GitHub-only accounts go 30 days after the end; a member
takes the register's moment by its own clock; nothing secret copied on
the device). The module README and Learn's AGENTS.md say the same.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
The reading of the full reading's fixes noted that macOS and Windows can
keep a window out of screen captures, which SecretScreen didn't ask.
PlatformScreenGuard now asks the app's window on Android, macOS and
Windows (asksOn), and both apps' runners answer app.halkyon/screen_guard,
counting what shows: macOS's MainFlutterWindow sets sharingType = .none,
Windows' FlutterWindow WDA_EXCLUDEFROMCAPTURE (Windows 10 2004 on;
before, WDA_MONITOR, black in captures). Neither runner was built here
(Linux): to be tried on each system.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
The reading of the full reading's fixes found that the device keeps
http://host/ and https://host/ in one place (folder, session key, keyring
prefix): someone with a stored http://host/ who added https://host/ and
then removed the first deleted the second's folder, session and keys while
it was in use, losing unsynced changes, and an anonymous account for good.
The places keep the names installs made (InstanceStorage.sharesWith says
which addresses share one): choosing an address the device keeps where a
known one is puts it in that one's place in the list, and removing an
address forgets nothing another known one uses. The build's own instance
keeps its place apart. Learn and Todo pass the storage's rule.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
The reading of the full reading's fixes found that a member, which keeps
no addresses, would delete everyone it holds with a GitHub ID as if
GitHub were their only way in (D1); that each deletion delivered every
pending notice, so with a member hanging the daily run grew quadratic,
and one failing person aborted it, forgetting no GitHub ID (D2); and that
a GITHUB_SIGN_IN_ENDS set wrong deleted data at the next daily run.
forgetGitHubAccounts now does nothing on a member and waits 30 days after
the end (gitHubAccountsKeptAfterEnd), keeping the accounts and IDs while a
wrong date can be put right; deletes each GitHub-only account on its own,
a failure logged by type, that person's GitHub ID kept for the next run;
and tells the members once, after all of them (deletePerson's tellMembers).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
The reading of the full reading's fixes found that every refusal at the
address's ceiling moved a reset to 72 hours after it (A1): someone who
knows only the address kept the owner's own reset from ever taking effect
with one burst every three days, and kept a forgotten one alive (A2).
Now a hold-out that begins before a reset takes effect moves it on only by
the time it holds the owner out and no hold-out counted before did
(account_reset.heldOutUntil, under a row lock), so the owner gets 72 hours
outside hold-outs in total; and every reset not carried out lapses 30 days
after it was asked (resetLivesAtMost), whatever moved it, besides a week
after it took effect.
Migration (module, Learn's and Todo's servers, Learn's device schema):
ALTER TABLE "account_reset" ADD COLUMN "heldOutUntil" timestamp without
time zone; nullable, existing rows counted from now on.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
ADR 0017's full-reading part says what the device's fixes did: the
clipboard, plain HTTP refused but on this machine, sheets never left busy,
what Save leaves, secrets out of debug logs, keyboards, screenshots and
backups, memory, and the real cryptography against the register's checks.
halkyon_app's README says what an app must add (its activity's screen
guard), Learn's AGENTS.md the HTTPS rule for instances.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
The recovery phrase, recovery codes and an authenticator's QR code and
key now show inside a SecretScreen: on Android it holds FLAG_SECURE while
it shows (through a channel each app's MainActivity answers, counting
holds, as several may show at once; ScreenGuard.kt), so screenshots,
screen recordings and the recents view don't get them; on iOS, which has
no such flag, it covers them while the app isn't in front, so the app
switcher's picture doesn't hold them (Dart-side, not checked on a device).
Desktops do neither: a window loses focus whenever its person turns to a
password manager. Todo's debug APK builds with it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
Equatable's own text lists every field, and a debug build's bloc observer
logs every state, so an authenticator's otpauth URI with its secret, the
recovery codes and the address went into the log. The password states
(signing in, getting back in, adding an authenticator, a step of the
account section) and AuthenticatorSetup now say only where they stand, as
the setup and new-phrase states did; Learn's observer takes a writer so
its test holds what it logs to that.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
ADR 0017's "Full reading (21 Oct 2026)": each server finding and what was
done (H1, M1, M2, M3, L1, L4, L5, L6, L7, L10), those accepted with their
reasons (L2, L3, L8, L9), and the device's M1, the owner's proof bound to
its operation and the whole base URL, still to do. "The rules as they
stand" brought up to date; the module README and Learn's AGENTS.md too.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
The full reading's M1 and L6: once GitHub IDs were forgotten at
GITHUB_SIGN_IN_ENDS, an account that had signed in with GitHub alone
counted as anonymous, so a stolen session and any mailbox gave it a first
password and authenticator; and such accounts, with no way in left, were
never pruned. Now:
- at the end, the daily run that forgets the GitHub IDs first deletes the
accounts that sign in with GitHub alone (no address, so never a
password), with everything they have, members told (deletePerson); the
app tells them for the three months before that their account goes then
unless they set up a password (passwordSectionGitHubEndsDeletes, en/de);
- password.set adds an address without GitHub only for someone anonymous
where the session lives (anonymousHere, which a member tells the
register, as it tells freshGitHubSignIn); anyone else is refused before
any code is used.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
The full reading's M3: whoever knows the password made the account's codes
wait, and that wait refused the phrase's owner their reset too, so they were
held out of both without end. Codes now count per account and per way they
came (SecondFactorPath: after the password, after the phrase), each with
its own waits; a code accepted either way clears both.
Migration (module, Learn's and Todo's servers; Learn's device schema
rebuilds the server-only table):
DROP INDEX "second_factor_failures_person_idx";
ALTER TABLE "second_factor_failures" ADD COLUMN "path" text NOT NULL
DEFAULT 'password'::text;
CREATE UNIQUE INDEX "second_factor_failures_person_idx" ON
"second_factor_failures" USING btree ("authUserId", "path");
Existing rows are the password's, unique per account, so unique per
account and way too.
Also the hosts' generated future calls, for the pruning's doc comment the
lapsed resets' commit changed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
The full reading's M2: a reset of the authenticators asked and forgotten
left the account to the password alone for good, and an account's reset
could be carried out a year after it was asked. Now a reset that took
effect and wasn't carried out lapses a week later, as never asked: the
password alone no longer signs in, a code still does, and asked again it
waits anew with a new mail. A reset of the authenticators carried out never
lapses (the password alone signs in until one is confirmed). The daily
pruning forgets lapsed rows (forgetLapsedResets).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
The tenth wrong owner proof in a row ends a session (the third
reading); someone holding that session tried the password or its
codes. The register now keeps the moment with the account's password
(`account_password.sessionEndedForProofsAt`): the account's sign-in
ways and its next password sign-in carry it, until the owner dismisses
it (`password.dismissSessionEndedWarning`, their session alone: it
hides a warning and secures nothing) or sets a new password. A member
ending its own session tells the register
(`noteSessionEndedForProofs`); a register out of reach leaves the
owner unwarned, the session ended all the same.
Migration (module, Learn's and Todo's servers, Learn's device schema):
ALTER TABLE "account_password" ADD COLUMN "sessionEndedForProofsAt"
timestamp without time zone; nullable, so every row keeps its values
and has no warning.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
ADR 0017 records what the fourth reading found and what was done, and
states who changes what of an account with which proofs, as a table the
full reading can check against; the module README and Learn's AGENTS.md
follow.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
The fourth reading's first finding: password.set took a session alone,
so whoever took one made a GitHub-only account, or an address account
without a password, theirs for good (their address or password, their
authenticator; GitHub then refused the owner).
- An account with an address proves it anew: a code mailed to that
address for this (EmailCodePurpose.firstPassword, its own mail), never
the session alone, a code asked for anything else, or another address.
- An account that signs in with GitHub alone does it within 10 minutes of
a GitHub sign-in (the session's method and creation time, where the
session lives; members tell the register), refused signInAgain before
any code is used; the address it adds is proved as before.
- Anonymous people still become identified with a code to the address
they add.
The app mails the code to the person's own address (asking for it where
the device doesn't know it), offers a password right after a GitHub
sign-in (D2's "at their next sign-in") before the person's app, and asks
for a GitHub sign-in anew when the instance refuses an older one, keeping
the code typed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
Codes by mail were tied to nothing but their address (the re-review's
seventh finding): a hostile instance could answer that a setup never
finished, have the person type a code into it, relay it to start the real
instance's account reset, and 72 hours later relay a second to finish it
with its own password and keys; the mail said "to sign in" whatever the
code was for. A code is now asked for a purpose (signUp, finishSetup,
authenticatorReset, accountReset, deletion), only that purpose takes it (any
other is refused as no longer valid, leaving the code as it was), and its
mail says what it would do ("to reset your account on learn.halkyon.eu: a
new password and new keys, and what was encrypted with the old ones is
lost") and to type it only in the app on that instance, nowhere else. The
app asks for each code with the purpose of its step.
Migration (module, Learn's and Todo's servers, Learn's device schema):
ALTER TABLE "email_code_request" ADD COLUMN "purpose" text; nullable,
so the requests of the last 15 minutes before it stay, taken by nothing
and forgotten within the hour.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
Wrong codes for removing an authenticator, new recovery codes or confirming
one counted toward the account's waits (the re-review's third finding): a
stolen session alone, without the password, could keep the owner's codes
waiting for up to a day at a time, renewed without end, holding them out
of signing in and of changing the password, the way to end that session,
and nobody was mailed. Those codes now count per session, where the session
is (the register's own, or a member's, which forwards the code): each try
counted before its code is checked (one statement, so tries sent at once
count each), a code accepted clearing the count, and the tenth wrong in a
row ending the session (sessionEnded; the device says it was signed out).
A try that checked no code (a refusal of something else) is taken back.
The account's waits now count only codes after a right password, and the
per-account ten tries an hour, which a session could fill as well, go.
Migration (module, Learn's and Todo's servers, Learn's device schema):
CREATE TABLE "session_code_failures" ("id" bigserial PRIMARY KEY,
"sessionId" uuid NOT NULL, "authUserId" uuid NOT NULL, "failures"
bigint NOT NULL, "updatedAt" timestamp NOT NULL) with a unique index on
"sessionId": a new table, nothing touched. Its rows go with a code
accepted, with the person, and hourly once their session is gone.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
Caddy's network had IPv4 alone, so Docker's userland proxy forwarded
every IPv6 visitor to Caddy from the network's gateway, and the server
counted all of them as one network (the second reading's M5, reasoned).
The edge network gets a unique local /64 (fd4a:6c6b:7e80::/64), Caddy a
fixed address in it beside its IPv4 one, both trusted as the proxy; the
CI runner's unit denies fd00::/8 and its sandbox check tries Caddy's IPv6
address too. A commit of its own: it needs Docker 27 or later with
ip6tables on the server and the network made anew (compose down, never
-v, then up), checked by hand first (NEXT.md).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
Whoever knew the password could guess codes at 10 an hour forever,
unnoticed, each guess checked against every authenticator (the second
reading's M4). Every code an account's second factor refuses after a
right password (sign-in, a phrase's reset, a change) or from a signed-in
device (removing an authenticator, new recovery codes, the code from one
confirmed before) now counts, per account, from every network; after
every tenth in a row no code is taken (codesWait, saying until when), an
hour, then twice as long each time, a day at the most (longestCodeWait,
second_factor_waits.dart): at a day, ten guesses a day still have about a
1 % chance a year against one authenticator, so the first wait mails the
owner that someone knows their password. The instance keeps no address,
so sign-in and the phrase's reset now send the address the device typed
(and its language), used only for that mail after its hash is checked
against the account's. A code accepted clears it all.
Migration (module, Learn's and Todo's servers, Learn's device schema):
CREATE TABLE "second_factor_failures" ("id" bigserial PRIMARY KEY,
"authUserId" uuid NOT NULL, "failures" bigint NOT NULL, "waits"
bigint NOT NULL, "waitUntil" timestamp, "updatedAt" timestamp NOT
NULL) with a unique index on "authUserId": a new table, nothing
touched.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
Anyone who knew an address could spend its whole hour: 10 sign-ins, 5
phrase proofs and 3 codes, each new code voiding the last (the second
reading's M3). Now:
- sign-ins begun count per address and network (10 an hour), phrase
proofs likewise (5), and networks as before (30);
- failed proofs count per address from every network: past 30 an hour
(wrong passwords, phrases' signatures by another key, an address nobody
has alike) no sign-in or proof begins for it anywhere, so guessing from
many networks makes at most 30 guesses an hour, while holding an owner
out takes failing without end from at least three; longer windows
would outlive the hourly pruning of the limits' notes;
- a signed-in change of password begins 10 times an hour per account;
- codes by mail count per address and network (3 an hour), per address
from every network (10, no mailbox flooded) and per network (10), and
up to 3 are valid at once, a fourth voiding the oldest, so codes others
ask for don't void the one its owner is typing.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
A key proof was a signature by the account's signing key, which every
signed-in device keeps in its keyring: whoever could read one, with a
code from its authenticator, set a new password without the password or
the phrase (the second reading's H1). The phrase now proves itself with
an Ed25519 key derived from its bits alone (halkyon_keys'
recoveryProofKeysOf, KDF context hlkrprf1, a sibling of the recovery
key), whose public half every setup, new phrase and account reset sends
and the register keeps (account_password.recoveryProofKey); the device
derives the secret half only while the phrase is typed. An account
without one (none exist outside development) proves no phrase until it
sets a new one; decoys check against a decoy proof key alike.
Migration (module, Learn's and Todo's servers, Learn's device schema):
ALTER TABLE "account_password" ADD COLUMN "recoveryProofKey" bytea;
nullable, so rows made before keep working and no data is touched.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
