| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Stage 8 of ~/next/improve-downloads.md, second half, plus the gap it exposed in
stage 7.
**Asking where to resume.** The node identifies an upload by (member,
directory, filename), so a client resuming one has to name the file — and
`transfer_open`, the obvious place to ask, travels in clear. Naming it there
would undo exactly what sealing this path bought in MNP 2.0: before it, the same
file was ciphertext leaving a node and plaintext arriving at one. So the
question is asked inside the seal that already exists, as an ordinary
`file_upload` with no bytes and `chunk_index: -1`. The node writes nothing,
creates no state, reserves no name, and answers with `resume_from` in the sealed
ack. A node that predates it refuses the index, which the client reads as "start
from the beginning" — the behaviour it had anyway — and the wait is bounded so
one that answers neither does not strand an upload.
The probe is answered after every check the write path makes, so it cannot ask
questions about a directory the caller may not write to, and it answers only
about the member who asks: otherwise one member could measure another's
progress on a file they never sent, and worse, resume it.
**Pausing an upload.** Reported: no pause button on an upload, even in the
desktop app. Stage 7 built pause around the download path — a target declares
whether it can be stopped — and an upload has no local target to ask. It was
also refused by design, since a transfer handed a lease it cannot re-create must
not be offered a button that would drop its slot for good. Uploads now ask for
their slot rather than being handed one, and say they are pausable outright: a
File is seekable and the node keeps the position. Resuming re-probes rather than
trusting the client's own memory, so it works across a reconnect too.
**And the slot they hold.** `_do_file_upload` never called `slots.touch(tr)`.
Chunks are not gated by the lease, so the file arrived — but the node reclaimed
a grant nobody appeared to be using after thirty seconds, twice, then abandoned
it, and the widget follows the lease. Measured from the journal: a 3.5 GB upload
read "waiting, 0 ahead" for a minute and a half while it was transferring. The
download twin of this was fixed on 2026-09-08; the same omission was still here,
invisible until uploads took a real lease.
`test_the_upload_itself_is_sealed` now checks every message `uploadFile` sends
rather than the first. Adding the probe put a second one in front of the one it
was written for, and it would have kept passing while guarding nothing.
Node suite 1202 passed, hub suite 850 passed.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HCGdheDLxGReuKHga3BtST
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Stage 8 of ~/next/improve-downloads.md, first half. Two defects that are the
same defect seen from two sides.
An upload's progress lived on the session, keyed by `rel_dir/filename`. A
dropped connection threw it away and the client's next chunk was refused with
`not_started`: an upload interrupted at 99% could only be started again from
zero, on a link flaky enough to have interrupted it once. It now lives in the
group context, keyed by member as well -- a shared directory means two people
can be sending IMG_1234.jpg at the same moment and neither may inherit, or
overwrite the position of, the other's.
What the lost state left behind was a `.part` nothing would ever finish, delete
or look at again. It is not an index entry, so it is invisible to every member
and to the operator's own file list: one abandoned film is a gigabyte of their
disk, kept for ever. That leak predates this branch.
A `.part` is deleted only when **both** hold: no upload is writing it, and
nothing has been written to it for 24 hours. Waiting costs disk; being wrong
costs somebody their upload, and is not reversible -- so a read-only root is
never walked (it cannot have received an upload), an unavailable one is never
walked (an unmounted drive reporting "nothing found" is how a careless janitor
deletes a library), and a file whose mtime is in the future is left alone (a
clock that went backwards is not evidence). The reaper matches whole paths and
the state records the path it is writing, rather than both sides rebuilding one
from a root name -- two implementations of one rule whose failure mode is
deleting a live upload.
The rules are in `uploads.py`, pure logic with no asyncio and no transport, the
same shape as `transfers.py` and for the same reason.
23 cases, four of them checked against the unfixed source. Node suite 1195
passed.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HCGdheDLxGReuKHga3BtST
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
"The node saw no transfer" was concluded twice from a journal that could not
have shown one: the open was logged at DEBUG, and the daemon runs at INFO. Two
diagnoses were built on that non-observation, and both were wrong.
Turning the root logger up to DEBUG is not the answer either — aiortc logs every
SCTP chunk, which on a 2 GB download is both unreadable and slow. One line per
transfer is not a volume problem, and it is the line that answers "did the
client ever ask for a slot, and what was it told".
Node suite 1172 passed.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HCGdheDLxGReuKHga3BtST
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Two defects in the lease machinery, both found in the node's own log, neither
reachable from any test on either side.
**`touch()` was never called.** The node ignored `tr` on `file_req` entirely, so
`used` stayed False for every download ever made and the sweeper revoked each
grant thirty seconds in — while the file was transferring at 20 MB/s. The pool
was correct and the handlers were correct; the call between them was missing,
which is why neither side's tests could see it.
**The requeue was a permanent cycle.** A revoked grant went back in the queue,
was granted again a millisecond later because there was room, and was revoked
again thirty seconds on. The node logged the same two reclaims every thirty
seconds for as long as it ran — minutes after the transfers involved had
finished. Three chances now, then the lease is closed and the peer told.
`test_the_counter_never_drifts` could not have caught it: nothing drifted, the
same lease simply never left.
A lease that starts being used forgets its earlier misses: a client that took
two grants to get going is slow, not abandoned.
Also `transfers show` reported the module defaults rather than the operator's
values until something had transferred, so `transfers set 2 2` answered
"applied now" and the next line said 0/8 — indistinguishable, from outside,
from the hot-swap that did nothing for months. The test asserted the defaults
and so agreed with the bug; found by typing the command.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HCGdheDLxGReuKHga3BtST
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Step 3 of ~/next/improve-downloads.md. Step 2 built the pools with constants;
this gives them to the operator, in the two scopes they belong to.
**The pools are the machine's.** `[node] max_concurrent_downloads` and
`max_concurrent_uploads`, default 8, on the §2.11 pattern: node.toml for a
fresh install, a roster.db override for immediate effect, editable from the
Node page and from `meshbay-node transfers show|set`, applied live through the
one `set_capacity` step 1 fixed.
**The per-member cap is a group's.** How many transfers one member may run at
once here — on the node like every other group setting (not the hub, which
would have authority over someone else's disk; not node.toml, which is
hand-written and needs a restart), changed by a signed operator instruction
(`OP_TRANSFER_LIMITS`, subject "d=2,u=2" so what is signed names the outcome),
broadcast to the group, and read live by the pools.
That was the one thing step 2's shape could not express: `per_member` was a
single node-wide number. `group_limits` and `member_cap(kind, member)` make it
a lookup — the group's own value if it has one, the node's default otherwise —
and it is deliberately the only dimension that is not node-wide.
Three refusals, each with a test:
- **absent means the default (2), never "unlimited".** A group that predates
the setting coming back unlimited would leave the node-wide pool as the only
control, which is the situation slots exist to end;
- **zero is not "unlimited"**, and is not "this member may not transfer"
either: the floor is one everywhere, and the CLI says to revoke the member
instead;
- **an unreadable row reads as unset**, not as zero — the same discipline the
sealed messages follow, where a payload that does not open must never become
a default state on its own.
`handshake_ack` carries this member's own caps for this group, so the interface
can say "2 of your 2 slots are busy" instead of drawing a bare spinner. Absent
reads as "no limit known" and the hint is not drawn — never as "unlimited",
which would have the interface contradicting the node.
1164 node, 793 hub, 0 failed.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HCGdheDLxGReuKHga3BtST
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Step 2 of ~/next/improve-downloads.md. A download is invisible to the node: it
is a series of independent `file_req` messages, with nothing saying one started
or ended, so there is nothing to count and nothing to cap. The lease is that
missing object.
`meshbay_node/transfers.py` holds the decisions and has no asyncio and no
transport in it, on purpose. The failure modes this has to survive — a slot the
node never gets back, a client waiting on a grant the node has forgotten — are
races through a DataChannel and unprovable there; here the clock is a parameter
and every method returns what changed, so the caller does the I/O and the tests
drive the worst case directly.
What it decides:
- two pools, downloads and uploads, separate from the stream pool: different
resources with different costs, and merging them makes both caps meaningless;
- per-member cap checked *before* the node-wide one, so a member at their own
limit queues behind their own transfers rather than holding a slot a second
member has none of. Per account across their devices, or the cap becomes a
function of how many tabs somebody opens;
- a queue that skips a member at their cap instead of waiting for them —
granting strictly in arrival order lets one member's limit stall everyone;
- `tr` drawn by the client and idempotent, which is what makes a reconnect safe;
- bounded per member, because unbounded queues are how a node runs out of
memory politely.
Every way a slot comes back, with the session teardown as the one that matters
(a closed tab, a quit browser and a dead network all arrive at
`shutdown_tasks`, and none of them needs a timer): explicit close, session
gone, a grant nobody took up in 30 s passed to the next in line, and a granted
transfer silent for 120 s reclaimed with its peer told, so a widget can offer a
resume rather than sit on a lie.
`GET /api/transfers` is the operator's window: when somebody reports a transfer
stuck at waiting, it is the only thing that says whether the node ever had them
in a queue — a log cannot, when the symptom is that nothing is happening. It
carries no filename and no path, which a test pins, because this is exactly
where one would be tempting.
Three things found while writing it, two of them mine:
- the randomised property test rejected `in_use <= cap` at once, and it was
right to: lowering a cap never interrupts a running transfer, so the count
legitimately sits above the new value. The invariant is that a *new* grant
never happens past the cap;
- the sweeper was started with `self._spawn`, which ties a task to one
session's set. It died with whichever peer opened the first transfer, and
every other peer's abandoned lease then stopped being reclaimed — a node that
fills up over days with nothing in the log. It belongs to the node now, with
its strong reference on the transport context;
- the pools are node-wide while `_peer_registry` is per group (finding H1), so
a slot freed in one group can grant one in another and the peer to notify is
not in the notifier's registry. Silently wrong in the first version.
Nothing enforces a lease yet: `file_req` is untouched, no client asks, and the
node grants everything. That is step 4's flag day, and this lands alone.
1148 node, 793 hub, 0 failed.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HCGdheDLxGReuKHga3BtST
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
`ops.set_node_settings` hot-swapped the stream pool by assigning
`webrtc._stream_sem`. That attribute has never existed on WebRTCTransport — the
pool is `ctx["_transcode_sem"]` — so `hasattr(webrtc, '_stream_sem')` was always
False and the branch never ran. The setting was accepted, written to roster.db
and node.toml, and applied only on the next restart, which is exactly what
draft-v6 §2.11 says it does not need. An operator lowering the cap on a
struggling machine, or raising it after "Server busy", saw nothing happen and
had no way to find out why.
`WebRTCTransport.set_capacity()` is the one implementation, on the object that
owns the state, so the download and upload caps the transfer-slots plan adds
next do not each grow their own copy of the mistake.
Resizing has semantics worth stating: the new cap governs new streams and never
interrupts one that is running, because a slot is held for the length of a film
and lowering a number must not take somebody's film away. The replacement pool
is built with the permits that remain (`new - in_flight`, floored at zero) — a
full set would briefly allow more concurrent viewers than either the old cap or
the new one.
That needs a count of slots in use, so `_stream_video` now maintains one instead
of the code reading the semaphore's private `_value`: a number this code keeps
itself survives the semaphore object being replaced underneath it, and the same
counter makes the "N of M in use" log lines mean something.
test_stream_capacity.py drives the real transport and the real `_stream_video`;
`test_ops_calls_the_real_mechanism` fails if the dead attribute comes back.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HCGdheDLxGReuKHga3BtST
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Chat messages have been signed by the sending device since MNP 2.0, but a
reader had no way to know that the device belonged to the account the node
named: the signature proved *a device*, and `sender_id` was still the node's
word. This closes that for any account a client has already seen.
**What was blocking it was not effort — the evidence was not being kept.**
`_do_device_add` verified the countersignature that admits a second device and
stored only `added_by_pk`: *which* key approved, never the proof. And
`device_add_transcript` binds `nonce_node`, the approving connection's handshake
nonce, so even a stored signature was unverifiable by anyone who had not been on
that connection. `identities` gains `add_sig`, `add_nonce` and `add_ts`, added
before the migration's early return — which fires on every roster widened since
2026-08-18, i.e. all of them, so putting them inside it would have meant they
never arrived.
`group_roster_req`/`resp` relays, sealed under a new groupbox purpose and
answered to **any member of the group**, every live device of every active
member with the evidence that admitted it. The node decides nothing: it hands
over evidence and the client walks the chain from each account's root outwards
(`_verifyRoster`). That is deliberate — the node is the party the property holds
against, so it is not asked to assert trust.
Two holes the tests caught while this was being built:
- "no signature" was being treated as a trust root, so a node that writes the
roster could put any key in an account's row and have it laundered straight
into the verified set. A root is a device that names **no** countersigner.
- pinning only the verified subset at first sight raised "key changed" on
legitimate second devices whose countersignature predates this change. First
sight pins everything the node says, because that is what trust-on-first-use
means and an alarm that fires on normal events stops being read.
The property, and it must not be rounded up: **once a client has seen an
account, a node that later substitutes a key for it is detected. Nothing is
gained at first sight**, where there is nothing to compare against — the same
boundary `per-node-identity-v1.md` draws, unmoved.
The cost, stated because it is real: the roster is member-visible, so every
member learns how many devices the others hold and their public keys. It stays
inside the group, the hub is not involved, and it is scoped per group. A member
who cannot see the keys cannot check them.
User-visible surface: one notice, "this account is using a key you have not seen
before", in ten languages. Nothing else.
16 tests — 7 on the node (the evidence is stored, it verifies from the roster
alone, a fabricated device carries none, another group's members are not
disclosed), 9 running the shipped `_verifyRoster` under node against rosters
built by the shipped Python: a chain of three in any order, a signature by the
wrong key, one for another node, one for another account, and two fabricated
devices signing each other admitting nothing.
Tier 3 (operator-signed roster attestation) stays deferred, with nothing
depending on it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TZZxYjz8YeWRz13xDi8LJr
|
| |\
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Both sides landed a breaking MNP change and both called it 2.0, which is right:
the sealed upload, the removal of `stream_seg` and mandatory chat encryption
share one flag day. They are recorded as one version in `__init__.py` rather
than as a race between two.
The resolutions that were decisions rather than mechanics:
* **`MNP_MIN_SUPPORTED` moves to "2.0".** The sealed upload alone was a
*confined* break — a 1.x peer could still connect, browse, download, stream
and chat, with only its uploads refused by `upload_not_sealed` — so the floor
deliberately stayed at "1.0". Mandatory chat encryption ends that
confinement: a 1.x peer can neither produce a sealed chat message nor read
one, so it would connect, look fine, and be unable to say anything. Refusing
it at the handshake is the honest form. The per-message `upload_not_sealed`
path is untouched and still right if the floor is ever lowered.
* **`sendChat` throws on an `error` reply**, from origin, applied to the sealed
send. It matters more after this change, not less: the node now refuses a
stale epoch, a malformed envelope and a device claim that is not the
connection's own, so there are three new ways for a message to be rejected
and none of them may look like a message that was sent.
* **`req_id` supersedes the per-type routing** this branch added for
`chat_keys_resp` and `device_hello_ack`. Both blocks are kept beside the
existing `chat_hist_resp` one, for the same stated reason — a node too old to
stamp — and their comments no longer claim to be the mechanism that closes
the class. `req_id` is.
* **`chat_send_probe.py` is rebuilt on origin's structure**, not beside it: two
scenarios, a stub that stamps `req_id`, `music_meta_req` as the older pending
request. The encrypted path is layered on — a real Ed25519 device key
generated in the page, and a `chat_keys_resp` sealed by the shipped Python,
because a payload the page built itself would prove only that the page agrees
with the page.
* **`test_reply_correlation.py` now sends a sealed message.** Its subject is
which of the two messages leaving that handler carries the id; plaintext chat
was only the fixture, and the node refuses one now.
* `groupbox` keeps both new purposes (`upload`, `chat_keys`); `protocol.py`
keeps origin's removal of `STREAM_SEGMENT` and this branch's correction of
the "Double Ratchet message" comment on `CHAT_MESSAGE`, which was wrong when
it was written and is wrong differently now.
Full suite on the merged tree: 1993 passed, 11 failed — the same 11 that fail
on a pristine checkout (2 Windows service tests, 1 apps-enabled policy, 7
transcode tests that pass in isolation, and the WebRTC invite test that hangs
on its own).
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TZZxYjz8YeWRz13xDi8LJr
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Downloads have been encrypted under a GEK-derived key since the beginning:
`file_chunk` and `stream_data` both go through `chunk_ciphertext`. Uploads
never were. `file_upload` carried the filename and the raw bytes in plain
msgpack, and `file_upload_ack` carried the name the node stored them under —
so the same file was ciphertext leaving a node and plaintext arriving at one.
There was no threat model behind that asymmetry.
Both halves now travel sealed under a third groupbox purpose,
HKDF(GEK, info="meshbay:upload:v1"). The filename, the destination folder and
the bytes are all inside the seal; only `upload_id` and `chunk_index` stay in
clear, because the node routes and orders on them before it can decrypt. This
direction seals *towards* the node — it holds the GEK for its own group — and
it opens the payload before it picks a destination or touches the disk.
What that forced, and why none of it is optional:
- `filename` was the correlation key on both sides. It cannot be: matching an
ack to its request by name would hand back exactly what the seal hides.
`upload_id` replaces it — client-drawn, opaque to the node, unique within a
connection, never an authorization input. The property it guarded (one
refusal fails one upload, not every upload in flight) is unchanged.
- Refusals can no longer quote what they refused. `No directory named 'X'`
becomes `No such directory in this group` plus the `code` that was already
there; the client knows what it sent.
- No plaintext fallback. A path that still accepts plaintext is not a sealed
path, so an unsealed `file_upload` is refused with `upload_not_sealed`.
Hardened while here, because what comes out of a seal is authenticated but not
validated — a member can seal anything: `filename` and `data` have their types
checked before any upload state is created, and `chunk_index`/`total_chunks`,
which are outside the seal by necessity, can no longer raise where a refusal
was meant.
Tests. `test_upload_sealed.py` pins the node half: nothing identifying on the
wire, tamper/wrong-key/wrong-group all refused with nothing written, and
multi-chunk reassembly unchanged. `test_upload_seal_client.py` drives the
shipped `uploadFile` over the shipped `crypto.js` under node and feeds its real
frames to the real `_do_file_upload` — the file lands intact, and the ack the
node actually produced comes back with the name it chose for a collision, which
is the half a source-reading test cannot see. Both upload purposes join the
JS/Python groupbox parity vectors.
BREAKING CHANGE: MNP 2.0. `file_upload`/`file_upload_ack` change shape on the
wire every deployed client speaks, which is MAJOR by the same rule 1.0 was —
but the break is confined to uploads. `MNP_MIN_SUPPORTED` stays at "1.0", so a
1.x peer still connects, browses, downloads, streams and chats; only its
uploads are refused, with a message saying which side is old. The client checks
the node's version before sending a chunk, so neither side meets this as a
timeout. This is the version negotiation shipped in 1.0 earning its keep: 1.0
cost a flag day, 2.0 costs a refusal code.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AsoWC3GmhNdwVFomW3QjH3
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
`stream_seg` answered with an MPEG-TS segment as base64 with no encryption at
all — the one message on the content plane that never went through a
GEK-derived key. Live on both transports, answering any authenticated member.
It predates `stream_data`, which does the same job properly (`chunk_ciphertext`,
keyed per segment, AES-256-GCM) and has since Phase 12. Its only browser caller,
`fetchStreamSegment`, was defined and never once invoked — a plaintext media
endpoint with no client. Removed rather than repaired.
Gone with it: `_extract_segment` and the ffmpeg semaphore in quic_server, the
`fetch_stream_segment` QUIC client method, and `_b64decode` in transport.js,
which had no other caller.
The H6 regression test lived on this handler — it pinned `_do_stream_segment_async`
to a coroutine so `subprocess.run` could not stall the event loop for thirty
seconds per request. It is replaced by the property that outlives the handler:
no transport carries media outside an AEAD, asserted on `stream_seg` and
`data_b64` across all three transport modules. The half of H6 that survives —
the live streaming path still spawns ffmpeg — keeps its own test.
BREAKING CHANGE: `stream_seg` is no longer answered on either transport. No
shipping client sends it. Recorded as part of MNP 2.0, whose other half — the
sealed upload — carries the version bump.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AsoWC3GmhNdwVFomW3QjH3
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
MNP carried no correlation id. A reply named its own type and nothing
else, so a client with more than one request in flight worked out which
one a message answered from the message itself — and for the replies
that name nothing it could not. `_dispatch` fell through to matching by
arrival order, which is a guess. `_sendAndWait` had the right value all
along: it keys `_pending` by `this._seqId++` and never put it on the
wire.
The guess fails asymmetrically, which is why it hid. The victim is not
the request that was answered wrongly — it is the unrelated one that now
waits out its own 30s timeout for a reply already delivered elsewhere.
Live on 2026-09-06: five `music_meta_req` sat pending for over 100
seconds behind a failing MusicBrainz, and a `device_list_result` was
handed to one of them. The composer is disabled while a send is in
flight, so a chat message whose reply went astray the same way left the
Chat tab looking frozen for thirty seconds, then unfroze on its own.
The `ack` half of this was fixed on 2026-08-30 by matching on request
type. That closed the instance and left the class open: a refusal has no
type to match on either, and `_dispatch_message`'s catch-all answers
every unforeseen failure with `{"type": "error", "detail": "Request
failed"}` — 238 of this module's 240 error sends name nothing at all.
`req_id` now rides on the request and comes back on the reply. On the
node it is published for the whole handler in a ContextVar and stamped
by `_send`: a parameter would have meant threading an argument through
all 240 send sites, and asyncio copies the context into a task, so a
handler that `_spawn`s its real work still answers under the right id.
It is never stamped on a broadcast — those answer nothing, and the
owner check in `_send` is what keeps a chat broadcast or an index push
from reaching another peer looking like a reply.
On the client, `_dispatch` resolves on `req_id` first and the
arrival-order fallback is gone the moment a node proves it stamps
(`_correlates`, armed by the handshake's own reply). The fallback stays
for an MNP 1.0 node, unchanged and no wider: there it is the only thing
there is, and removing it would leave device_list_result, join_result
and the handshake replies reaching nobody.
Two things fall out. `sendChat` refuses an `error` reply like every
other request in the file — it returned it as success, which did not
matter while a refusal reached the wrong caller anyway and would now
show a rejected message as sent. And `_group_ctx` uses `.get`: a reload
pops a removed group while sessions connected to it are open, and every
request they had left raised KeyError into that same catch-all.
Sealed index messages are the one exception to the fast path. They
cannot be handed over until they are opened, which is asynchronous while
`_dispatch` is not — resolving on the id alone gave `fetchIndex` the
envelope and skipped `onIndexSync` entirely. Caught by extending
`index_seal_probe.mjs` to stamp a reply the way a current node does,
after the hub suite passed over it: the probe built its own frames and
had never seen one.
Tests, all failing before and passing after: `test_chat_send.py` drives
the real ChatPanel over the real transport for both shapes of reply with
an older request pending (3 of its 6 are new, and the 3 for `ack` pass
either way, so it discriminates); `test_reply_correlation.py` pins the
node's half — the refusals that name nothing else, the broadcast that
must not be stamped, and a late reply from a spawned task answering
under its own id rather than the most recent request's.
Full suite: 1897 passed, same 11 pre-existing failures as before.
QUIC keeps its own dispatch and is not stamped. It is disabled by
default and no browser request reaches it, but the asymmetry is real.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Dn1xYx9uT69mCB6UDvyKAN
|
| |/
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Chat messages are sealed with AES-256-GCM under a key derived per group, per
epoch, per *device*, and signed over the ciphertext with the device key the
node pinned. The node relays and archives; it cannot read a message.
There is no switch. MNP goes to 2.0 and MNP_MIN_SUPPORTED moves with it, so a
1.x peer is refused at the handshake with `version_too_old` rather than
admitted and then unable to speak. An opt-in flag was designed and rejected:
every node is a test node, so it would have bought nothing and left a plaintext
branch reachable — C6's lesson one feature later. A test reads the source and
refuses any code that consults a `chat_encrypted` setting.
Not Sender Keys, and `senderkeys.py` is now documented as unused. With
distribution under the group key and a node that serves history to devices
which were not present, the node must retain each chain's earliest key, and a
chain key at iteration i yields every message key from i on by pure HKDF —
forward secrecy is zero either way. What the ratchet was left buying was
stateful client code with silent failure modes, three of them reproduced: any
member could sign as any other, a second device dropped the first's chain, and
the skipped-key cache grew without bound. The reasoning is in
docs/chat-sender-keys.md, which is the specification and the decision record.
Epochs, not rotation: the epoch key is wrapped under the group key at delivery
and never stored under it, so `gek_rotate` is a re-wrap. A group-key-derived
archive key would have made every message ever sent unreadable on the first
`member unpin`, which is the documented step after removing a member. A new
epoch opens on member revoke/unpin, device revoke and `gek_rotate`; old epochs
are kept and still delivered, so history stays readable to everyone who could
already read it, and nothing anywhere deletes one.
Three prerequisites this needed, each a live defect on its own:
* The peer registry was keyed by user_id, so one account's second device
evicted the first and the broadcast skipped recipients by account — a
person's phone never saw what they typed on their laptop.
* The handshake authenticated an account, never a device. `device_hello`
(additive, signed, refused unless the key is a live device of this account in
the node's own roster) is what lets the node refuse a member claiming
somebody else's key.
* `_admin_exec_file_delete` authorized against the exact uploading key, so
device linking had already broken deleting your own file from your other
device. It now authorizes against any non-revoked device of `uploader_id`.
Found by driving the real panel over the real transport, not by reading source:
`chat_keys_resp` was routed by arrival order and handed to an unanswered
`media_meta_req` — the original frozen-tab defect in a message type that did
not exist when that probe was written. And `_asText` had been deleted with an
unrelated helper beside it; its only caller sits inside a promise the panel
catches, so every conversation rendered empty with nothing in the console.
Existing node data is migrated by QE/migration/migrate_chat_encryption.py
(not versioned, per the QE rule), run with the node stopped.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TZZxYjz8YeWRz13xDi8LJr
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Caught in review, and the review was right. The previous commit documented the
flag as deprecated so that `--help` and the man page would agree. That solved
the wrong problem: the flag contradicts the model this whole refactor exists to
establish, and the coherent answer was to delete it.
It wrote `upload_dir` into a *brand-new* `[[groups]]` block, and
`GroupConfig.__post_init__` reads that key by forcing every other root
read-only and appending that path as the one writable one. So
`group add --dir X --writable --upload-dir Y` silently made X read-only — two
mechanisms deciding which directories accept uploads, one of them invisible, in
a group created after the model that replaced it.
Gone from the CLI, from `ops.attach_group`, from the loopback API and from the
MNP `group_attach` payload, which now carries `writable` instead. The *read*
path in `config.py` is deliberately untouched: an existing node.toml using
`upload_dir` must keep working, and that is the only legitimate use left. The
man page says so under the config key, and no longer lists an option.
The test that guarded the deprecation wording now guards its absence — and
earned itself immediately by finding a `group add` usage string still offering
the flag.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011pvMdvLBG92jyhvD5pD6us
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Every other test of the plugin architecture reads source for the *absence* of
app names. That proves nobody wrote a special case for Videos; it cannot prove
a genuinely new application works, because there was no new application.
This is one. It stores directories, appears as a tab, has a settings pane and
lists files, and the node has never heard its name outside a single allow-list
entry. Two files and one registry line, which is the claim
`docs/refactor-groups.md` §4.1 makes.
It ships hidden behind `?dev=1` (`dev: true` in the registry, the same opt-in
shape as transport.js's `?trace=1`). Registering it normally would put a toy
app in every operator's group; not registering it would prove nothing, since
registration is exactly what is claimed to be sufficient.
**Adding it found two places where the claim was nearly true rather than true,
and both are fixed by making the code less app-specific:**
`group-settings.js` fell back to the whole registry when a group had no
`enabled_apps` yet — which would have turned a hidden app on for everyone. It
asks `availableApps()` now.
`group-page.js` wrote out `videoDirectories` / `musicDirectories` /
`photoDirectories` by hand, so a fifth app would have needed that file edited.
It derives `<key>Directories` from the registry.
Neither was found by reading; both were found by adding the app, which is the
whole reason it exists.
Verified in a real Electron window as well as by the tests: hidden by default,
present with the flag, offered its own settings section, and listing exactly
the files under its configured folder and its subfolders — not the ones beside
it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011pvMdvLBG92jyhvD5pD6us
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The control was hidden and its `canCreateDir` left computed and unused. It is
back in the Files toolbar as an icon: the toolbar already carries one labelled
primary action, and a second beside it competes for the width the breadcrumb
trail needs. The name is in `title` *and* `aria-label` — a title is invisible
to a screen reader on a button with no text, so an icon-only control without
both is simply unnamed for anyone not reading with their eyes.
Its gate changes. It required `isNodeAdmin`, which contradicted the node's own
rule — "making a directory is not a privileged act; a member who can add a file
can organise where it goes" — and hid the control from everyone who could have
used it. It now follows the Upload button: a writable root, and not at the top
of a group, where the level is the set of roots rather than a directory on
anyone's disk.
Restoring it surfaced a real gap. `_do_dir_create` never learned about RO/RW:
`_do_file_upload` gained the `writable` check with the model and this one did
not, so a member refused a file in a published library could still leave empty
directories all through it, and could write to a drive mid-eject. Read-only has
to mean read-only for every way of writing, not just for files.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011pvMdvLBG92jyhvD5pD6us
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The directory table travelled on `index_sync` alone — a *full* index, which the
node only ever sends on request. Every ongoing change went out as an
`index_delta`, which carried files and nothing else. So the message that says
"something changed" was the one message that could not say a root had.
The acks hid it: `root_add_ack` and friends broadcast the new table to whoever
is connected, so the common cases looked right. What that could not cover was
the operator's own client, where the ack landed and was then overwritten — the
table calls `onRefreshIndex` after an add, that fetch returns the set from
*before* the node's reload (fire-and-forget, because a rescan is minutes on a
real library), and `applyIndex` writes it over what the ack had just delivered.
The new directory appeared for one paint and vanished.
Two halves. `index_delta` now carries the roots table, sealed with the rest and
identical to `index_sync`'s — additive, so a 1.0 client sees a field it does not
read. And the table no longer refreshes the index after a root change: the ack
gives it the new set immediately, and the delta the node pushes when the scan
finishes gives it again, along with the files.
The test that pins it uses an *eject* as its case, because an eject changes no
file at all — the entries freeze — so its delta is empty of additions,
deletions and updates. Without the table it says literally nothing, which is
how a library disappearing from under a group went unannounced to everyone but
whoever pressed the button.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011pvMdvLBG92jyhvD5pD6us
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
There is no `uploads/` subdirectory any more, and the client names the folder
rather than the root.
It was the last of v5's quarantine — the per-user layer went on 2026-08-14 for
the same reason — and it goes on the same grounds: a folder appearing beside
the operator's library because somebody sent a file is the node deciding how
their disk is arranged. Somebody dropping a file into the folder they are
looking at expects it to be in that folder.
**What made the quarantine worth having was never the subdirectory.** It is the
filename allowlist, the size cap, the chunk ordering and the no-overwrite rule,
and all four are untouched: an existing file is never replaced, the second
sender of IMG_1234.jpg gets a free name, and the check still sits at the write.
Letting the client choose the destination is safe for one reason and only one:
it is resolved through `RootSet.resolve()`, which refuses `..`, absolute
segments and anything whose resolved form escapes its root, symlinks included.
A member answers "which of this group's folders", never "which path on the
operator's disk" — and the test that used to assert the node chose now asserts
that, with six shapes of escape.
`direct` goes with it. Its only job was to say "no subdirectory for this root",
which is now every root, and a config flag that does nothing is worse than none.
Chat's attachment folder finally does something: the directory the operator
picks in the Chat settings pane is where attachments are written, falling back
to the first writable root while they have not chosen one, or if the one they
chose has since been made read-only or ejected — a stale choice should not
become a refusal at send time.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011pvMdvLBG92jyhvD5pD6us
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The previous commit was the wrong fix. `add_root` did leave the running node
unchanged, but editing the live `RootSet` in place — which is what I did — is
wrong in the other direction.
`DirectoryIndexer.retarget` decides what to scan by diffing the names it
already holds against the ones it is handed, and `_retarget_indexer` hands it
`groups_ctx[gid]["roots"]`: the very object the op had edited. So the new root
sat on both sides of the comparison, nothing was scanned, and the directory
would have appeared in the table permanently empty. `_reload_config_inner`
diffs the same way and would have concluded nothing changed. `remove_root` had
the same shape and would have kept serving a removed directory's files.
The real defect is that two front doors did different things. `ui/app.py` has
always fired the daemon's `reload_fn` after these ops, which re-reads node.toml
and builds a *fresh* set; the MNP path retargeted a stale object instead. That
asymmetry is exactly what `ops.py` exists to prevent, and it is why the bug
survived until an operator added a directory from a browser — the loopback path
worked all along.
So: the ops leave the live set alone, `_retarget_indexer` asks the daemon to
reload, and `update_root` keeps editing in place because flags change no files
and the synchronous upload handler reads that object on the next request.
The tests now check the files rather than `describe()`, which proves nothing
about whether anything was scanned. One of them demonstrates the failure mode
instead of describing it, so the rule is checkable and will say so if
`retarget` ever changes. Two more cover the seam itself — that the MNP path
reloads, and that a context with no daemon still retargets.
Diagnosed by reading the running node's journal rather than the source: the
first add logged "Reloading config" and a rescan, the two later ones logged
neither. I should have looked there before the first attempt.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011pvMdvLBG92jyhvD5pD6us
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
`video_root` (a string), `audio_root` (a string) and `photo_roots` (a list)
said the same thing three ways: three roster accessors, three ops, three MNP
messages, three admin-op subjects. They become `set_app_directories(app_key,
paths)` and its single-directory wrapper, stored under `<app>_directories` and
keyed by the app's registry name — so an application can be added without
touching this layer, which is the whole claim of the plugin architecture.
The three old names still work. Their MNP messages are handled, and the roster
falls back to the old key when the new one is unset, so a node upgraded into
this keeps working with no migration step — the plan called for a script, and
a script nobody runs on the machine where it matters is worse than a fallback.
Two things are new rather than moved:
The paths are validated. The setters this replaces accepted anything, so a
typo — or a path left behind when a root was removed — was stored happily and
then matched no entry, leaving an app showing an empty tab with nothing to
distinguish "misconfigured" from "no files yet". Deliberately not
`RootSet.resolve()`: that also refuses a currently-unavailable root, and an
operator must be able to point an app at a library on a drive they ejected.
The legacy scalar is derived, never stored. `video_root` still rides on the
handshake ack for MNP 1.0 clients; kept as a second stored value it would
drift from the list within one run, which reads as "it works after a restart".
Also here: chat's own two settings (a directory, which must be on a read-write
root because it is a destination rather than a view, and a link-preview switch
gating the unfurl path — checked before the cache, or turning it off would
still serve every preview already fetched), the `app_directories`,
`chat_directory` and `chat_link_preview` MNP messages, the plural
`<app>_directories` on the handshake ack, and `music` as the app's one
identifier where storage said `audio` and the registry said `music`.
The Music enricher now resolves a boundary per configured directory rather
than one for the group: with several, a single boundary is wrong for all but
one of them, and for Music that is the difference between reading a folder as
an artist and reading it as a release.
Suite: 11 failures, all pre-existing.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011pvMdvLBG92jyhvD5pD6us
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Review of the Phase 1 commit found the RO/RW model sound but three paths
unfinished, each of which broke the flow the phase exists to deliver. Plus
29 test failures it introduced and no coverage for anything it added.
Uploads went to the wrong directory. The node read a `root` field on
file_upload that no client ever sent, so every upload landed in the first
writable root while the Files toolbar offered its button based on the root
being browsed — with two writable roots, uploading from one wrote into the
other. Files now names the root it is showing; Chat names one chosen in the
shell (an operator-configured directory arrives in Phase 2); the node refuses
an unknown name rather than falling back, and refuses read-only and ejected
roots by code.
Shared directories were unreachable on the web. The table read its roots
only from the loopback API, which resolves to "not available" in a browser,
so the section rendered for nobody there — while the Uploads controls it
replaced had worked — and the transport.updateRoot/ejectRoot/plugRoot methods
beside it were dead. MNP is now the path, loopback the fallback for a local
node with no live connection, and adding a root over MNP takes a typed path
since no web page can browse a remote disk.
Ejecting updated nobody's screen. transport.js resolves an admin ack against
the pending request and returns, which is right for every op whose caller
knows the value it chose; the root acks carry state only the node can compute,
so the operator who clicked Eject was the one client that never saw it happen.
And the ejected flag reached roster.db but was never read back, so a restart
undid it and the next scan read an empty mount point as an erased library.
Also: the member-upload endpoint answered 200 and did nothing (removed); the
wizard ignored the first root's RW switch; reload compared roots on name and
path, so editing writable in node.toml did nothing; the table had no path
column, which is the only thing separating two libraries sharing a basename;
apps_enabled normalisation differed between the two sides of a signed subject.
Tests: eject/plug, per-root upload refusal and the node.toml rewrite had no
coverage at all. test_member_upload_policy.py is replaced by
test_root_writable_policy.py — it tested a removed feature — and every
property worth keeping from it moved rather than being dropped.
Docs: draft-v6 structural decision 9 is annotated as superseded (the operator
can no longer have a directory only they may write to — a real capability
removed, flagged rather than hidden), the man page documents the root verb and
the RO/RW fields, and refactor-groups.md §7b records what the plan got wrong.
Suite: 41 failures before, 13 after — all 13 pre-existing on main.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011pvMdvLBG92jyhvD5pD6us
|
| |
|
|
|
|
|
|
|
|
|
|
| |
Replace the upload boolean with per-root writable/removable/ejected flags.
Backend: new ops (update_root, eject_root, plug_root), MNP 1.1 protocol
messages, live RootSet updates so API always reflects current state, CLI
root subcommand (add/remove/set/list/eject/plug).
Frontend: SharedDirectoriesTable with optimistic toggle switches, eject/plug
in Files and Settings, upload gated on root.writable, ejected-root filtering
in all media apps, updated Create Group wizard, 10-locale i18n.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
| |
"DataChannel closed" from a peer and a clean node log look identical: the
answer-ready line reported only the srflx count, not the host addresses.
On a NAT'd host or a VM the sole host candidate is an address no other
machine can route to, and that is exactly the case you cannot see. The
line now reads `... host: 192.168.200.173, 1 srflx`, so "did the node
offer anything routable" is answerable from the journal.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
`ice_interfaces` compared the operator's entry against ifaddr's
`adapter.name` only -- the kernel name on Linux (`wlp3s0f0`), but the adapter
GUID on Windows (`{846EE342-...}`). A setting written on Linux, or copied
into a Windows node's node.toml, matched no adapter at all.
The failure was silent and total rather than partial: aioice binds one socket
per host address, so an empty list means no sockets, no host candidates, and
an SDP offering only a reflexive address. The settings field is free text
with no picker, and on Windows the operator sees neither the GUID nor the
description -- `ipconfig` shows the connection name -- so an entry now
matches the adapter name, the device description, or one of the adapter's own
IPv4 addresses, case-insensitively. An address is the one identifier visible
on every platform.
A filter that matches nothing now falls back to the unfiltered list with a
warning. Losing the 5 s timeout saving is a regression; being silently
unconnectable is a defect.
Also fixes IPv4/IPv6 discrimination in the same loop: the two were told apart
by falling through to an `elif` that index-probed `ip.ip[0]` and `ip.ip[2]`,
which on an IPv4 str yields characters that compared unequal by luck rather
than by design. Now discriminated by isinstance.
WINDOWS-PORT.md claimed Transport had "no platform dependency"; it does.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DtfG7z6wHWj8RKHCvxQtY1
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
node.toml, the keystore envelope, the unlock key, the loopback UI token,
pairing/invite code files and the denylist were all read and written with
the platform default encoding and newline translation. On Windows that is
cp1252 + CRLF: a node.toml or keystore holding any non-ASCII byte failed to
load, and ops.py's line-based node.toml editor round-tripped CRLF in and
LF out.
Every read is now `encoding="utf-8"`; every write is `encoding="utf-8",
newline="\n"` so the files stay LF whatever the OS. No-op where the locale
was already UTF-8.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
| |
Platform directories, signal handling, chmod guards, ffmpeg discovery,
and platform-conditional CLI messages — all testable on Linux.
See docs/WINDOWS-PORT.md §5 for the plan these implement.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
`index_sync`, `index_delta` and the `handshake_ack` config payload now travel
sealed under a GEK-derived subkey (`meshbay_common/groupbox.py`, mirrored by
`sealGroup`/`openGroup` in `crypto.js`). Only `type`, `v`, `group_id` and the
ack's `node_pk`/`proof`/`sig` stay in clear — a receiver must route and
authenticate before it would trust a decryption. Verify, then decrypt.
The ack line is integrity, not confidentiality: the signed handshake transcript
names no ack field, so `is_node_admin`, `enabled_apps`, `video_root` and the
rest were authenticated by the DTLS channel alone. The index line is defence in
depth against a repeat of C1/C6 — a peer served before the handshake completes
now gets ciphertext, not filenames. Nothing against an observer, the hub, or a
member; that is the whole claim. `index_progress` stays clear (D3, counters
only). Chat is out of scope.
Failure is fatal: a payload that does not open ends the session naming the
message type — never an empty index or an empty `enabled_apps`, both of which
are legitimate states.
Version negotiation ships here too (phase 15.6, brought forward): `v` + `v_min`
on `handshake` and `handshake_challenge`, refused with `version_too_old` /
`version_too_new` / `version_unreadable`. The flag day was already being paid
for; the next breaking change now costs a refusal message.
BREAKING CHANGE: breaks the WebRTC wire every deployed client speaks. Hub and
every node must deploy together; the SPA is served by the hub, so a browser
picks up the new client on reload. See MESHBAY_NODE_PROTOCOL.md §11.1a, §13.1.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HkzbhmMmK8PqQBtGz5zCvY
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
`file_chunk` and `index_sync` were each built twice, once per transport, and
the two copies did not agree. WebRTC sent binary, unsigned chunks carrying a
`file_id`; QUIC sent base64 fields, two BLAKE3 hashes, a per-chunk Ed25519
signature and no `file_id`. `index_sync` was plain entries on one transport
and a `GroupIndex.serialize()` envelope on the other. One message type, two
shapes, one consumer each, and nothing that failed when they drifted — finding
C6 one size down, in the two places the handshake unification did not reach.
Phase 9.15 moved WebRTC to the binary format and dropped the per-chunk
signature; the QUIC encoder was never brought along. It is dropped here rather
than reintroduced: the AES-GCM tag authenticates the ciphertext under a
GEK-derived key, and since C3 the node authenticates itself once in the
handshake instead of once per megabyte.
`meshbay_common.protocol` now owns the chunk codec (`chunk_ciphertext`,
`file_chunk_wire`, `file_chunk_plaintext`) and `meshbay_node/transport/wire.py`
the index builder, which also absorbs the delta the daemon used to hand-build.
`test_transport_wire_parity.py` fails if either server grows its own copy back.
`ChunkRequest`/`ChunkResponse` are deleted. `ChunkResponse` described the QUIC
half while reading like the contract for both, which is what made the fork hard
to see at all.
BREAKING CHANGE: MNP 0.15 changes the encoding of `file_chunk` and `index_sync`
on the QUIC transport. The WebRTC shapes are byte for byte unchanged and no
QUIC client ships, which is why this is a MINOR bump; a deployed QUIC peer
would have made it MAJOR.
Also fixes a test fixture that put a `Path` where the daemon puts a `RootSet`.
Nothing caught it: the old QUIC index handler never touched `roots`, and
`entry_abs_path` fell through `Path.resolve(strict=...)`, reading the virtual
path as a truthy flag and returning the right file by accident.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AsoWC3GmhNdwVFomW3QjH3
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Opening a different season of the same show moved everything under the
synopsis, which is where the season control and the episode list are, so the
thing just clicked was no longer under the pointer.
- The synopsis is exactly three lines for a multi-season show, with a "read
more" link floated into the third line box (-webkit-line-clamp only ever
puts its ellipsis at the end of the last line and leaves no room after it).
Clamped from above and pinned from below to the same number: a constant,
not a range — a season summary runs two lines and the next one twelve, and
a band still reads as a jump. Whether three lines is all of it depends on
the modal's width, so it is measured in the browser and re-measured on a
resize.
- The cast is clamped to two lines.
- SeasonMenu replaces SeasonTabs: the tab row scrolled sideways once a show
had more seasons than fit, which is close to unusable on a phone. One
trigger reading "Season 5 · 1997" and a menu of every season with its
episode count, one row high whatever the season count.
- media_meta_resp.director was filled from the credits crew's job ==
"Director", a movie shape. TMDB's aggregate tv_credits crew is routinely
empty and never carries that job, so every show answered null and the modal
dropped the line. It now comes from created_by on the show details. Cached
show metadata keeps its null until TMDB_META_TTL_SECS expires or an
operator re-matches.
The facts line is joined rather than concatenated (a title with no rating
used to open with " · ") and carries the show's own year next to the
director; the selected season's air year moved onto the picker.
test_video_detail_measured.py asserts rectangles through layout_probe.py, not
declarations: the picker's offset inside its own modal body is the same pixel
either way, the synopsis and cast heights, where the read-more link lands,
and the open menu at 320 px. Each measured block sits in a whole-pixel-height
container, or two identical layouts an eighth of a pixel apart round to tops
one pixel apart. test_tmdb_show_director.py covers the credit.
docs/mediacenter.md §10.4.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014UtzVrzM7e2tG9fSpkR9ML
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The link-preview fetch is an outbound request to an address a member
chose. safe_url() already blocked non-public addresses and re-checked
each redirect hop; this adds the parts that were missing:
- Rate limit. `_do_link_preview_request` was reachable by any member
with no ceiling, so a member — or a hub minting tokens for many
accounts — could drive unbounded outbound HTTP from the operator's
machine (amplification / DoS / on-demand IP disclosure to arbitrary
hosts). Now bounded per connection (15) and node-wide (60) over a
60 s window; only a real fetch counts, a cache hit is free, and over
the ceiling the reply is a plain `ok: false` (bare link), not cached.
- Port allowlist. safe_url() passed `parts.port` straight through, so
a member could aim the node at `http://<public-host>:<any-port>`.
Restricted to {80, 443, 8080, 8443} — every real OpenGraph page,
none of SSH / mail / DB / cache / search / admin ports.
- DNS rebinding. The connection's actual peer address is now
re-checked against the public-address rule (`_reject_if_rebound`),
so a name that resolves clean and then to something internal does
not get its body read. Best-effort (no `network_stream` extension,
no check); a full literal-pin is noted as remaining hardening.
- Decompression bomb. `_downscale` now refuses an image whose header
dimensions exceed ~40 MP before convert()/thumbnail() decode it.
Third security review, finding M3.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011pG75yGK3NthNfyjH74omG
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The unified handshake reached QUIC in Phase 11.5, but the chat and
stream handlers did not get the authorization rules the WebRTC path
gained at the same time:
- _do_chat_message_sync took `sender_id` from the wire, so an
authenticated peer could post as anyone (NS6 / M2a). It is now the
authenticated session's id, always.
- chat used a connection-global peer registry and read chat_store from
the top-level context, so on a multi-group node a message from one
group fanned out to peers of the others (M2b / H1). Both are now
resolved per group via _peer_registry() / _group_ctx(), mirroring
the WebRTC path. The QUIC peer set is kept separate from the WebRTC
one in the same group context — the two session types have
different _send signatures and no cross-transport fan-out is wired.
- _do_stream_segment_sync ran `subprocess.run(timeout=30)` on the
event loop with no concurrency cap, so one request stalled the whole
node and any member could fork-bomb it with ffmpeg (M2c). Extraction
now runs in a thread behind a small semaphore, spawned as a tracked
task (cancelled on connection_lost).
Also corrects the stale docstring claiming C6 is still open here — the
GEK proof has been enforced on this transport since 11.5.
No behaviour change for shipping clients: none speak QUIC, and the
listener is off by default (previous commit).
Third security review, finding M2.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011pG75yGK3NthNfyjH74omG
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Two bugs fixed:
1. `meshbay-node member invite <user>` created a local roster invite
but never told the hub to add the user to group_members, so the
group was invisible in the SPA. The node now calls
POST /v1/groups/{id}/members/{username} after creating the invite,
and the hub endpoint accepts node-scoped tokens (the admin_id
check is the real authorization guard).
2. The WebRTC handshake let a previously-pinned user reconnect without
a code even when a new invite was pending (e.g. after leave + re-invite).
Now any pending invite forces code entry, regardless of existing
member/pin status.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
aiortc's connection_kwargs() keeps only the first STUN URI from
RTCConfiguration.iceServers ("only a single STUN server is supported"),
and aioice.ice.Connection has a single stun_server field. So the node's
four default STUN servers -- and anything added on the Node page or with
`meshbay-node stun add` -- collapsed to stun:stun.l.google.com:19302. When
that one server was slow or unreachable from the node, ICE gathering
(get_component_candidates, timeout=5) burned its full 5 s with no
server-reflexive candidate, adding seconds to every browser connection.
The multi-server fallback of draft-v6 s2.12 was configuration only.
transport/stun_multi patches aioice.ice.server_reflexive_candidate (same
monkey-patch technique ice_filter.py uses on get_host_addresses) so a
single ICE gather races the STUN binding request against every configured
server on the one bound socket and takes the first answer. One reachable
server anywhere in the list now yields a reflexive candidate in one RTT.
- daemon: install_stun_multi() alongside install_ice_filter()
- ops.set_node_settings: push the list to stun_multi.set_servers() so the
CLI / Node-page hot-swap takes effect without a restart
- webrtc_server.handle_offer: log ICE gather time and srflx count
- test_stun_multi.py: fan-out, first-answer-wins, all-fail, empty-list
fallback, DNS failure
Verified end to end with a real RTCPeerConnection: with a black-hole STUN
server first in the list, gathering still completes in ~0.07 s with full
srflx candidates (previously a 5 s stall).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BSsQhfxEAhwi4nqc4hASmq
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The passphrase derives two independent client-side values: auth_key (the
hub verifier) and bundle_key (AES-GCM key for the per-node identity
bundles, which live on nodes and never on the hub). Changing or
recovering a passphrase is therefore two operations — swap the hub
verifier, and re-wrap every reachable node's identity bundle.
Flow A — change a known passphrase (Profile page)
- POST /v1/users/password re-proves the current passphrase, swaps
pw_hash/salt/version, revokes every refresh token and returns a fresh
pair so the tab that made the change stays signed in.
- MeshBayTransport.rewrapAllNodes: for every group's online node, connect
with the old key, read the identity off the handshake, store it back
under the new key. Returns updated / unreachable / failed so the UI can
point at the operator-unpin fallback for the gaps. Always-shown
confirmation dialog listing reachable and unreachable groups.
Recovery key
- keyderive.js generateRecoveryKey (32 random bytes, grouped Base32) and
deriveRecoveryKey (HKDF-SHA256, domain meshbay:recovery:v1:<username>).
- Every per-node identity gets a second copy wrapped under the recovery
key: keypair_bundles.bundle_enc_recovery (node-only column, added in
_SCHEMA_KEYPAIR and via a PRAGMA-guarded ALTER for existing DBs),
carried on keypair_bundle_store / _resp. MNP 0.13 -> 0.14, additive.
- session.recoveryKey is persisted in IndexedDB (slot rk) and lazy-loaded
on connect, so a group joined in any later session still leaves a
recovery copy.
- Shown once at registration; optionally folded into the verification
e-mail as a pass-through the hub never stores or logs, with an opt-out.
- Profile -> Recovery key re-loads R and backfills every reachable node
via rewrapAllNodes in bundleKey mode (no passphrase re-entry).
Flow B — recover a lost passphrase (#/reset, linked from sign-in)
- POST /v1/users/password/reset-request {username, email}: both must be
the pair on file, checked against the blind email_hash (never
decrypted). A mismatch — wrong e-mail, unknown username, non-active
account — takes the identical no-op path (no code, no mail, same 200),
so it reveals nothing and cannot be used to spray reset mail from a
username alone. 5/min, 1-hour single-use code.
- POST /v1/users/password/reset {username, code, new_auth_key}: same
expiry / attempts / single-use checks as e-mail verification; revokes
every session and deletes every registered device key so a stored one
cannot sign back in past the reset.
- ResetPasswordPage: request code -> code + optional recovery key + new
passphrase -> reset + sign-in -> fan-out. connect() falls back to the
recovery-wrapped copy when the passphrase key cannot open bundle_enc.
Without a recovery key: sign-in is restored and each group needs the
operator-unpin fallback.
Supporting fixes (found in live testing)
- member unpin now also deletes the keypair bundle; connect() mints a
fresh identity when handed a bundle it cannot open (unless _rewrapOnly,
set by rewrapAllNodes), so a rejoin completes instead of dead-ending
before the invite-code prompt.
- A browser with no bundle key gets a passphrase prompt on the group page
instead of a "go back to the browser you registered on" message.
- RegisterPage / LoginPage / ResetPasswordPage trim the username so every
key derivation matches the hub's stored form.
Docs: docs/auth-confirm.md. Locale keys across all ten catalogues.
Tests: test_password_change, test_password_reset, test_recovery_email,
test_recovery_key, test_rewrap_fanout, test_bundle_store_recovery, plus
additions to test_admin_ops_mnp and test_webrtc_transport. Hub suite 492
passed; node suite 741 passed (the lone test_packaging_units failure is a
pre-existing RPM-spec flake, reproducible on main).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GGkxJW9br8Y9bhT8ywJ3oc
|
| |
|
|
|
|
|
|
|
|
| |
The WebRTC transport relied on a single Google STUN server — if it was
unreachable, ICE gathering waited the full 4s timeout. Now four public
servers are used by default (Google ×2, Cloudflare, Mozilla), configurable
via node.toml, the Node page UI, and the CLI (meshbay-node stun list|add|
remove|reset). Changes are hot-swapped on the live transport.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Expose invite_ttl_hours, pair_ttl_hours, device_request_ttl_minutes,
max_concurrent_streams and transcode_incompatible_video in the Node
management panel. Changes are applied immediately via roster.db and
written back to node.toml so they survive a DB wipe. On startup,
roster overrides take precedence over node.toml defaults.
Draft v6 §2.11 documents the design; MNP gains node_settings_set /
node_settings_set_ack for the browser path.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Root of the 2026-08-29 demo35 storms. `_do_media_meta_request` could not
tell "this is a movie" from "this video isn't enriched yet" — both have
season/episode None — so during a slow initial scan with a browser on the
Videos tab, every show episode requested was run through the *movie*
search path with its raw filename as the query
(`search/movie?query=Show S01E01 1080p WEB DL ...`), hundreds per second,
until TMDB rate-limited and posters stopped loading. Worse, an
un-enriched show episode's own valid cached "tv" match was treated as
stale (its provisional kind was "movie"), so a file already resolved got
re-queried anyway.
- While a video is un-enriched (no display_title — enrich.py always sets
one), never run a TMDB *search*. Serve the cached match if the content
hash has one, honouring the cached kind ("tv"/"movie") rather than the
provisional split; otherwise answer confidence 0.
- Once enriched, the strict `cached_media_type == media_type` check is
unchanged: an enrichment fix that reclassifies a folder movie->tv still
drops the stale match and re-resolves.
- video-app.js: `useMediaMeta` gains an `enrichSig` dependency
(`entry.display_title`) so the client refetches once the enriched fields
arrive on an index delta — the fileId is a content hash and never
changes, so nothing else would retrigger it.
Not caused by the V8-V13 work; it raised the per-file call count so the
pre-existing race became a visible storm.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018BMLQjqFGCize2KtNBT79v
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
A one-click alternative to the full "Fix match" search-and-pick flow, and
reachable without SSH (`meshbay-node video rematch` clears a whole group).
- MNP 0.13: tmdb_rematch / tmdb_rematch_ack (additive — an older node
logs "unknown type", the button just does nothing). OP_TMDB_REMATCH,
signed like tmdb_override (media_cache is shared node-wide).
- media_cache.drop_tmdb_match(file_id): forgets the match AND the
override marker — deliberately stronger than clear_file_tmdb, since the
operator is explicitly asking for a fresh resolution.
- webrtc_server: _do_tmdb_rematch / _admin_exec_tmdb_rematch, dispatch +
admin-response routing, broadcasts tmdb_rematch_ack.
- transport.js: rematchTmdbMatch(fileId, signFn); 'tmdb_rematch' in the
admin-op allowlist; tmdb_rematch_ack handled like tmdb_override_ack.
- video-app.js: a "Re-match" button beside "Fix match" in the detail
modal (isNodeAdmin), then bumpMediaMetaGeneration(). video.rematch_one
key in all ten locales.
docs/mediacenter.md §10.1: V8–V13 marked done.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018BMLQjqFGCize2KtNBT79v
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
sequel_variants
V8: the TV/show branch of _tmdb_search used the old "first candidate over
0.6 wins" shape. It now shares one _tmdb_ladder helper with the movie
branch — score every candidate query, keep the best, fast-path a
confident primary hit. A year lifted off the show's folder name
(title_parse.year_in, e.g. "Some.Show.2022.S01") rescues a sub-0.6 hit
that lands on the exact year. title_parse.clean_query de-dots a
folder-derived title without naive_title's extension-stripping trap.
V9: _best_match gains an optional `year`. When the top result is not a
confident textual hit (ratio < 0.6) and a year was requested, a
different result of that exact release year is preferred — TMDB already
year-filtered the search, so this is a hard corroboration, not the fuzzy
re-rank §3.3 warns against. A confident top hit is never overridden.
search_movie/search_tv forward the year.
V10: sequel_variants widened — trailing Roman→digit as well as
digit→Roman, spelled-out indices (one..twelve / un..douze / ordinals),
and a "Part N" / "Chapitre N" wrapper. Still empty for a trailing word
that is not an index or a 4-digit year.
V11: when the primary hit is already decent (>= 0.6) and there is nothing
more specific to try (no alternative_title, no sequel variant — only a
punctuation restatement left), the ladder returns without the extra
requests. The clean-title common case is back to one call.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018BMLQjqFGCize2KtNBT79v
|
| |\ |
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
A batch of wrong poster-grid matches found live on a real library
(2026-08-29): a two-volume film's second part matched the first; a
numbered sequel matched a same-year making-of documentary; several
entries of one franchise matched a single early entry whose localized
TMDB title is the franchise name; one matched nothing. One mechanism:
_tmdb_search returned the first candidate query whose title-similarity
ratio merely cleared 0.6, before alternative_title / the Roman-numeral
variant was ever tried.
Matching:
- title_parse: fold guessit's volume/part number back into display_title
so the parts of a multi-part film stay distinct in the query, the card
and the override.
- _tmdb_search: keep a strong PASS 1 fast path (ratio >= 0.85, one
request), otherwise score every candidate query and pick the best. A
year-exact rescue lifts a sub-0.6 top hit to the confidence floor only
when TMDB's own year-filtered result lands exactly on the filename's
year. No local re-ranking of any single result list; no tmdb.py change.
Fix match / rematch:
- _admin_exec_tmdb_override: a movie override touches its own file only
(guessit gives a whole franchise one display_title); a show override
still fans out. Corrected files are marked in media_cache.tmdb_override.
- media_cache: tmdb_override table; clear_file_tmdb / clear_tmdb_matches
drop auto-resolved matches while sparing manual corrections.
- ops.rematch_video + `meshbay-node video rematch` (loopback endpoint +
CLI verb): re-resolve a group's video matches after a matcher fix.
file_tmdb is keyed by content hash and otherwise only pruned on
deletion, so nothing dislodged a cached match before.
- a rename now drops the stale auto match too (daemon
_reenrich_renamed_video_entries).
UI:
- VideoDetailModal shows the source filename and resolved TMDB id; an
unmatched poster gets a badge (3 new video.* i18n keys x 10 locales).
So a wrong match can actually be identified before hitting Fix match.
docs/mediacenter.md 10.1 records this and the V8-V13 follow-up backlog
(show-branch ladder, year-aware _best_match, wider sequel_variants, the
0.6-0.85 extra calls, movie grid merge, per-card rematch).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018BMLQjqFGCize2KtNBT79v
|
| |/
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
VPN/virtual adapters
aioice sends STUN binding requests from every IPv4 interface and waits up
to 5 seconds for all to complete. On a machine with Tailscale (wt0), the
STUN request never gets a response, adding a fixed 5-second penalty to
every WebRTC connection — measured at 6 s total (vs 1-2 s without it).
Auto-exclude virtual/VPN adapters (tailscale, virbr, docker, veth, podman,
cni) and CGNAT-range IPs (100.64.0.0/10). Operator can override with
ice_interfaces in node.toml [node] section for explicit control.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Wizard (Electron):
- Auto-provisions node config (hub URL + username) from logged-in user
- node:start handles both cold start and restart of misconfigured daemon
- Waits for daemon to reach 'running', auto-links node key on hub
- probeNode accepts intermediate states for wizard progress feedback
Reset (meshbay-node reset):
- Unlinks node key from hub (DELETE /me/node_key, best-effort)
- Stops and disables daemon (systemctl --user disable --now)
- Erases ~/.config/meshbay, ~/.local/share/meshbay, ~/.local/state/meshbay
MusicBrainz contact:
- Resolved from owner's hub email instead of per-node roster config
- Removed musicbrainz_contact UI and WebRTC handshake field
- Removed set_musicbrainz_contact/musicbrainz_contact from roster
Node pairing:
- Added operator pairing banner on NodePage
- Added operator_paired flag to list_groups
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Regression from device linking (Stage C, 2026-08-18). Once a device is
pinned on a node — as a member of one group, or an operator pairing — the
`known` fast-path in `_do_join_request` dropped straight into `_join_ok`.
For any *other* invite-only group it had no roster row for, that answered
`not_authorized_for_group` and stopped there: the client never got
`code_required`, so the pairing-code form never appeared and a legitimately
invited member could not join.
The `known` branch now, when there is no membership for the group being
opened:
- with a valid code → consumes the invite and admits (as the unknown-
device path already does);
- with no code but an invite waiting for this user here → `code_required`,
so the client prompts;
- with no code and nothing inviting them → `not_authorized_for_group`,
unchanged, so the H3 guarantee (a hub-invented pin gets no key) holds.
Also fixed: the group's own roster row is now consulted first, so an
existing member opening their group is never mistaken for a stranger.
Tests in test_roster_pairing.py cover all three branches plus the H3 guard.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018gKJ85aZyvEwarXMFzFEwi
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Paste an http(s) link in a group's chat and it unfurls into an OpenGraph
card — title, description, site name, and image — the way WhatsApp/Signal/
Slack do it.
The fetch is the node's, never the browser's or the hub's. The browser
cannot: a strict img-src/connect-src and CORS block it, and a direct fetch
would leak every reader's IP to the linked host on each render. The hub must
not touch group content (draft-v6 §2.5). The node already fetches third-party
metadata for the Videos and Music apps, over the same authorised path.
Flow mirrors media_meta_req: the client sends `link_preview_req {url}`, the
node replies `link_preview_resp` with the card fields (or `ok: false`), and
any OG image is stored under its blake3 in the existing media_cache thumb
store — the client then fetches it via the normal file_req path, exactly like
a poster. Nothing durable is added: the card text lives in a bounded in-memory
TTL cache on the node (draft-v6 §2.7 — enrichment on demand, the asking device
caches), and MNP goes 0.11 → 0.12 (additive: an older node logs "unknown type"
and the client shows the bare link).
Because the URL is chosen by a *member* and triggers an outbound request from
the operator's machine, `linkpreview.safe_url` is an SSRF gate: http(s) only,
no credentials, and every resolved address must be globally routable — no
loopback, private, link-local, multicast or reserved range, cloud-metadata
included. Redirects are followed by hand so each hop is re-checked. Residual,
documented in the module: DNS rebinding between the check and connect, closed
properly by pinning the checked IP — a follow-up.
Also fixes a long-standing chat annoyance the preview cards made worse:
opening the Chat tab landed a screen or two above the newest message because
the scroll-to-bottom ran before attachment thumbnails and (now) preview cards
had loaded and grown the content. A ResizeObserver keeps the view pinned to
the bottom through late content growth, and does nothing once the reader
scrolls up.
Tests: test_linkpreview.py (the SSRF gate and the OpenGraph parse, incl.
redirect re-validation and image downscaling) and test_link_preview_request.py
(reply shape, the media_cache image round-trip, the result cache).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018gKJ85aZyvEwarXMFzFEwi
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Found live: "fix match" appeared to work for two shows but not for a movie
whose own automatic search kept landing on the same wrong result. The
override only ever recorded the file->tmdb_id mapping — never the
metadata that id actually names. _do_media_meta_request's cache check
agrees the mapping is fresh (same media_type) but finds nothing under
that *new* id in tmdb_meta, since nothing had ever fetched it, and falls
through to a brand-new search using the file's own title — reproducing
the exact match the override was meant to replace.
This stayed invisible for the two shows only because their own title
happened to be enough for that fallback search to land on the right
answer anyway, entirely independent of whatever the override recorded —
never because the override was actually being honored. It surfaced on a
movie whose own title search kept landing on the same wrong match
regardless.
Fetches and stores the real metadata for the chosen tmdb_id up front (via
the existing _tmdb_build_meta, which needs only the id — no search result
object required), so a later lookup finds the override itself instead of
falling through to a search blind to it.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The real reason a node restart alone didn't fix already-indexed entries
after the previous commit's enrichment change: _do_media_meta_request
checked media_cache's file->tmdb mapping (keyed by content hash) and
trusted it unconditionally, before ever comparing it against the file's
*current* movie/show classification. A file whose season/episode changed
on a later scan — exactly what the Specials-folder fix does, for every
file it reclassifies from "movie" to "tv" — kept answering with its
stale, wrong-kind-of-match forever, since nothing about a reclassification
touches this cache or its key.
Now falls through to a fresh search whenever the cached media_type
disagrees with what the entry resolves to right now, rather than trusting
a mapping that predates the file's current classification.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Client (transport.js): a localStorage ring buffer of connection/ICE/
DataChannel state transitions, visibility changes, request timeouts, and
periodic health pings — enabled once via ?trace=1 (persists), read back at
any time via #mb-debug without devtools. Off by default, zero behavior
change unless enabled.
Node (webrtc_server.py): MESHBAY_WEBRTC_TRACE=1 gates ICE-state-change
logging and a per-session heartbeat (message count, seconds since last
message, ICE/connection state) every 30s.
Debugging aid for the "stuck after several minutes of mobile screen lock"
report — not a fix. Stays on this branch until confirmed useful/resolved.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
IndexEntry.path is the *folder* a file is in (indexer.py's
_virtual_dir docstring: "the directory a file appears in"), not the
file itself. GroupIndex.get_entry_by_path() treated it as if it named
one file, and every one of its four callers did too:
_do_music_meta_request, _do_media_meta_request, _do_tmdb_override, and
_admin_exec_tmdb_override. Any two files sharing a folder — an album is
one folder with many tracks, a season is one folder with many episodes
— collided: a lookup by path silently returned whichever entry the
index happened to iterate to first, regardless of which file the
client actually asked about.
Found live (2026-08-25): three unrelated albums ("High Tone - Various",
two "Le Peuple de l'Herbe" albums) all showed the same MusicBrainz
cover, because all their representative tracks happened to sit in one
"high_tone" folder alongside a track that legitimately matched that
cover. A force-reload didn't help — the bug is server-side, not a
stale client state.
Fixed by keying these four request/response pairs by `file_id` (the
entry's own content hash — already unique, already how every other
lookup in the system identifies a file) instead of `path`, both in the
wire messages (music_meta_req/resp, media_meta_req/resp, tmdb_override)
and in music-app.js/video-app.js's own hooks. GroupIndex.get_entry_by_path
is now unused and removed — GroupIndex.get_entry(file_id) already did
the right thing.
No test previously exercised either handler with two entries sharing a
folder — the only existing coverage (test_tmdb_override_policy.py) gave
each entry its own folder, so the bug never had a chance to show up.
Added that scenario there and in two new test files, all confirmed
failing against the pre-fix code before being confirmed green against
the fix.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013XSohfUQQiaE77qyFLgSv3
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
A new group application (docs/apps.md's plug-in mechanism), following the
plan in docs/photos.md. Unlike Videos/Music: several photo roots per group
instead of one (photo_roots is a set, one signed op replaces it whole),
a single album-grid view with no third-party matching step, and per-photo
info read from the file's own EXIF at index time — no metadata service,
no credential, no outbound network call at all.
Protocol (meshbay-common, MNP 0.10 -> 0.11, additive): `taken_at`/`camera`
on IndexEntry; `photo_roots`/`photo_roots_ack`; `OP_PHOTO_ROOTS`.
Node: roster.py stores photo_roots as a group_settings entry (JSON list,
same shape as enabled_apps); ops.py/webrtc_server.py validate and sign the
whole set in one op, same pattern as apps_enabled; a new PhotoEnricher
(indexer/enrich_photo.py) runs Pillow in its own small bounded pool,
separate from the video/audio pools, producing a resized thumbnail plus
the two EXIF fields — never GPS, checked by a grep-based regression test.
Client: photos-app.js — one album card per directory containing images,
a per-album photo grid, and a lightbox with next/previous (keyboard and
buttons), zoom in/out/fit/100% starting from the actual on-screen fit
percentage, and a "zip this album" button reusing files-app.js's own zip
mechanism (lifted into file-utils.js's downloadDirectory so both call the
same implementation). group-settings.js gets an add/remove multi-root
picker, distinct from Videos/Music's single-value one.
Bugs found and fixed before this ever shipped, worth keeping the story of:
- enrich_photo.py read width/height from the raw image *before* applying
EXIF orientation correction, and read DateTimeOriginal off the plain
0th-IFD Exif object — a real camera stores it in the Exif sub-IFD, which
Pillow only exposes via get_ifd(Exif). A flat, hand-built EXIF dict
round-trips through Pillow either way, which is exactly what would have
hidden both bugs; the regression test builds EXIF with piexif instead,
matching what real hardware produces.
- photos-app.js's album grouping stripped a trailing path segment from
entry.path under the assumption it still carried a filename — it
doesn't (files-app.js's own convention: e.path is already the
containing directory), so every album collapsed one level into its
parent. Found live against a real multi-folder library.
- transport.js's ADMIN_OP_TYPES allowlist (already the fix for an
identical bug on video_root/apps_enabled, see 4783d81) was missing
photo_roots: its admin_challenge matched no pending request and was
silently dropped, so saving a photo root just timed out after 30s with
no error.
- daemon.py pruned a thumbnail when its file left the index (root removed
or reconfigured) but never forgot the content hash was "already
attempted" — the same bytes reappearing under a renamed/relocated root
(an operator's real workflow) were then permanently skipped, forever,
with nothing to indicate why. Discarding the attempt alongside the
cache entry on prune is what makes pruning actually reversible.
- packages/meshbay-client's app:// protocol handler served every file
with no Cache-Control header, so Chromium was free to serve a stale
cached copy indefinitely — none of several `npm run sync-ui` + reload
cycles during development actually picked up the new code until the
renderer's disk cache was cleared by hand. Now sends Cache-Control:
no-store.
- the lightbox's zoomed image used flex centering (align-items/
justify-content: center) combined with overflow: auto — a well-known
trap where the browser centers overflowing content by shifting it, and
the leading half of that overflow (here, the top of a zoomed photo)
sits outside what the scrollport can actually reach. Reported live as
"unusable". Fixed by switching to top/left alignment once zoomed.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TiZG4AuSnxHohQMpwTHTyL
|