| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
A snapshot leaving out earlier epochs' metadata (`snapshot::left_out`) and a rejoining
node (`catchup::prepare`) read it from the document as it stands. It is never writable
through the store, and under a policy the rules take a change to it only from an
epoch's first; but with no policy, screening judges only an epoch's root, so a member's
own change could rewrite or remove an entry: a replay window narrowed or widened, a
translation or a clock changed, a strict remake's entry gone.
Now both read it at the epoch's snapshot change (`snapshot::as_built`), as
`verify_document` reads the epoch's own: a snapshot from its base's epoch's, a
rejoining node from the target's. Every earlier entry is a copy in that change, so what
was written there since doesn't count. The first epoch follows none and has none.
|
| |
|
|
|
|
|
| |
An import uploaded as `application/octet-stream` takes `rules`, `baseHeads`, `settle`
and `switchDelay` in the query, but not the retentions: for those the document had to
go base64-encoded in the JSON form. Now `replayWindow` and `archiveRetention` go in the
query too, as JSON like the rules (`{"epochs":1,"age":"7d"}`, as in the body).
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Every snapshot carried the metadata of all earlier epochs (`_continuum/epochs`)
forward, so it grew by an entry, a clock per actor among it, with every remake. Only a
rejoining node reads an earlier entry: a node on epoch P reads P+1's (its clock, when
it was proposed, its replay window) and every one up to the swarm's epoch (kind,
translation, whether it was strict). It replays only while P+1's replay window covers
it; beyond, its writes are quarantined whatever the entries say.
So a snapshot now leaves out the generations older than the oldest whose replay window
still covers a node on the epoch before it (`snapshot::left_out`), judged by the
remake's `proposedAt`: the same on every node, and earlier than any rejoining node's
clock, so nothing a node can still replay through goes. Metadata that doesn't read
stays, with all after it.
A strict remake's metadata stays for good: by it a rejoining node, however far behind,
tells a strict remake of a swarm strict already (`catchup::prepare`, which reads the
generations it finds). There is at most one, short of such a remake. No new key and no
change to the rejoin: a node beyond the window that finds some entries and not others
quarantines with "cannot translate" rather than "outside the replay window".
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Review 3, R3: 1186 kept honest nodes from proposing or voting for a
strict remake of a swarm strict already, but a majority of members,
each with keys of its own, could still commit one, and switch and
catchup followed it without asking what kind of remake made the
epoch: the registry, policy and bans of a strict swarm changing by a
member majority, the honest minority following.
A node now refuses to switch to such an epoch (its document's policy
strict, the remake's kind strict), and catchup refuses to rejoin one
whose chain of epochs holds a strict remake after the swarm was strict
(as the old document is, or once a remake since made it so). access.md
says so.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
With the policy's reads kept, judging a change still costs about one
read (some 3.5 µs), and so does judging without a policy: what is left
is At::new, which asks get_change_by_hash for each head. Automerge
builds the change again from the document's operations to answer, at
a cost growing with its history. Screening asks it again for every
candidate and every dependency, and sync for heads.
Ask by the missing dependencies instead (changes::holds), which look
the hash up: one the document holds is not among them. A change queued
in the document, waiting for its dependencies, isn't either; but then
some are missing, and only then is it asked the slow way. Every check
of whether a document holds a change goes through it.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Review 2, N7, the care it raised: a joining node's empty document took
as the first epoch's root whatever dependency-less change came first,
from whichever peer it joined through; the first epoch's ID is all
zeros, so nothing pinned it. A remake's epoch has its ID, the snapshot
change's hash, which the rules hold its root to.
* Every node names its epoch's root in HELLO and HELLO_REPLY (the
epoch's ID for a remake's, its genesis change for the first; wire
version 45). A node whose document starts from another root of the
same epoch refuses the handshake, logging it: another history under
the swarm's ID, a forged root or a genesis started again.
* A joining node with an empty document takes as its root only the one
it was given (--genesis-root, new), else the one its peers name, and
none while they name different ones (access::screen::RootPin, kept
in Screening across documents): one peer hands it another root only
if it is all the joiner hears, or never, with --genesis-root. A root
refused is listed with the refused, what builds on it with it.
* The genesis node logs the root (`The swarm's root`); GET
/admin/epochs shows it (`root`).
A genesis started again under a remembered swarm ID used to merge with
the nodes of it still running; they now refuse it (another root), and
the docs say to join them with seeds instead.
The epoch's identity stays as it is (generation 0 all zeros): records
and epoch metadata name it so, and every comparison of epochs with
them.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Review 2, N2: nothing checked that a strict remake's base wasn't strict
already. In a strict swarm the registry, the policy and the bans change
only as a user who may change them signs; a strict remake committed by
a majority of members, each with keys of its own, could drop any of
them (unban, remove managers, empty userSigned and the sealed
declarations), and the honest minority would follow.
A node now takes a strict remake (proposing or voting) only in a swarm
not strict yet: turning strict is done once, by a remake; after it, by
signatures.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Review 2, N3: 1150 let any member write an abort decision once a remake
had expired a grace period ago, judged by the change's own time, the
writer's word. A member could race the coordinator's legitimate commit
with an abort dated as it liked; both taken, Automerge shows
one, and nodes that acted on the commit split from those that didn't.
The two writers of such aborts are operator requests whose endpoints
already need write on _continuum/remakes: aborting an expired remake
(DELETE /admin/remakes/{id}) and superseding one (POST
/admin/remakes?force=supersede). Both now commit as the user who asks
(the proposal and its vote with them, for a supersede), which the rules
take as one who may write the decision. The clause judging an abort by
its time goes, and with it the time the rules carried. An expired
remake is still aborted through any member, by an operator.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Review 2, N1: a strict manifest wasn't bound to its base. Any member
could propose an old, legitimately signed one again at the current
heads; nodes would verify it and vote, and every value at a signed
place changed or added since (a ban, a new user, a rotated key, a
changed userSigned or sealed declaration) would be deleted, nothing
after the base being left to replay.
The manifest now names the epoch and the heads it was made at, and a
node takes it (proposing or voting) only for a remake of that epoch
with exactly that base. GET /admin/strict-manifest makes it so;
continuum-signed strict proposes at those heads already.
|
| |
|
|
|
|
|
|
|
|
|
|
| |
Review 2, N12: a strict remake kept only the values at the signed
places its manifest names, and copied what holds them as it was. A
member that replaced _continuum/users by a scalar before the turn left
the strict epoch without a registry, where any member may do anything.
The holders of the registry, `_continuum/access` and the bans are now
made anew as maps, with only what the manifest keeps copied into them;
and a manifest keeping no user is refused (at proposal, vote and build
alike), as a strict swarm without one would trust every member.
|
| |
|
|
|
|
|
|
|
| |
Review 2, N10: the judgement before a change leaves (1140) and a
replay's (1141) skipped changes when the document *with* the change had
no policy it enforces. A write removing the policy, or setting one the
node doesn't know, then left unjudged, while receivers judge it at its
dependencies, where the policy was enforced. Both now read the policy
at the heads before the change, as receivers do.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Review 2, N4: replay_newest names the node the document's author once,
before its first group. Taking a refused group back forks the document
at the heads before it, and Automerge's fork has no author; when the
refused group was the actor's first change, the next one was committed
as seq 1 without the footer naming the node, refused as "it names no
node", taken back too, and so on: the whole replay in the quarantine
with a misleading reason.
The fork now names the node its author again. Test: the first group
refused (a capability revoked in the new epoch), the next replayed.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
A strict remake takes a majority of members whose --strict-keys take
its manifest's signer: no node votes for one without, so
--follow-strict-remakes (1173) goes. A node that doesn't vote still
switches once the others commit, as with any remake.
So that an operator can tell why one fails: a strict remake aborted
for want of a majority (the voting period over, or a majority out of
reach) says in its reason (GET /admin/remakes/{id}, and the
coordinator's log) which members didn't vote, and that a member votes
for one only with its manifest signed by a key of its --strict-keys.
Each of those logs why itself (`Not voting for a strict remake`).
The sim a_strict_remake_too_few_take: only a takes the manifest; the
remake aborts, naming b and c.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Until a swarm is strict, the user a change names is its node's word, so
any member could have planted registry entries, policy, bans or
user-signed values as any user, plain or signed by a key it put there
itself; writing the strict policy trusted all of it as it was.
A remake of a new kind, `strict`, turns a swarm strict instead:
* Its content is a manifest (access::strict): the swarm, the strict
policy, and each value kept at the places values are signed, by the
SHA-256 of its JSON, signed by an operator (ssh-keygen -Y sign -n
continuum-strict).
* Every node builds the next epoch as compaction does, keeping at the
signed places only the values the manifest names, as they were at
the base, and writing its policy (RemakeKind::protocol, the one kind
that changes protocol data). Writes after the base reach the new
epoch only by replays, judged by the strict policy.
* A node votes for one only with its manifest signed by one of
--strict-keys, keys given out of the document; without any, for none
(logged once, `Not voting for a strict remake`), unless
--follow-strict-remakes. A coordinator proposes only what it would
vote for (403). Once a majority commits one, every node switches, as
with any remake: refusing to switch would only cut a node off.
* GET /admin/strict-manifest gives the manifest keeping everything held
now, the values for review and the heads they are at;
continuum-signed strict shows them, signs once confirmed and proposes
at those heads.
The sim a_swarm_turning_strict_through_a_remake: a member plants a
registry entry as a manager; the operator's manifest leaves it out; a
and b, given the key, commit; c, without, follows; the entry is gone
and the policy strict everywhere.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Review finding 7, second part: a record carried into a new epoch whole
(proposal and decision in one change) was taken from any member, for
any remake ID. A member could write a forged record, decision included,
for a fresh ID in a single change.
The snapshot's epoch metadata now names the remake's coordinator, as
the quorum built it. The carried-record exception holds only in a
later epoch, only for the remake that made it, and only from the
coordinator its snapshot names, where there is no record yet. Wire
version 43: the snapshot, and so the epoch ID, changes.
The review suggested carrying the record inside the snapshot instead.
The snapshot can't hold the decision, which names the snapshot's own
hash. So the snapshot vouches for who may carry the record instead.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
A switching or rejoining node replays late writes as fresh changes
naming each write's user, and nothing judged them before they left: a
replay every node would refuse (a revoked capability, a second
envelope at a signed path, readers changed) was refused everywhere,
and with it everything the node wrote after it.
Under a policy, replay_newest now judges each group it commits as
every node will (rules::judge_held, in the new epoch); one they would
refuse is taken back (forked at the heads before it, same actor) and
its paths quarantined with the reason. replay_newest takes the epoch.
Review finding 3 (remake replays).
|
| |
|
|
|
|
|
|
|
|
| |
POST /admin/quarantine/{id}/replay needs write on _continuum, which
never covers user data, and replayed every path of the entry as the
user who asked: every node refused the user data in it. It now
replays only the paths the user may write and keeps the rest in the
entry (left), for one who may; 403 if the user may write none.
Review finding 3 (quarantine replays).
|
| |
|
|
|
|
|
|
|
| |
A node refusing changes a peer brought tells it which broke a rule; the
peer judges them itself, and if its rules refuse them too, drops them
with all built on them (its document forked without them, a fresh
actor), quarantining its own writes among them. Wire version 42.
Auth plan M6 (rejoin on taint).
|
| |
|
|
|
|
|
|
| |
A remake's replay writes each value as the user who wrote it,
splitting where several users wrote under a path, so replays are
judged as the original writes.
Auth plan M6 (remakes, 2.4).
|
| |
|
|
|
|
|
| |
Forget records and released quarantines name the user who made them,
so they are judged by its capabilities.
Auth plan M6.
|
| |
|
|
|
|
|
|
| |
registry::exists counted any value at _continuum/users as a registry;
keys_at answers None where there is no map, and its users (records,
members, forget records, remakes) say what they make of it.
Auth plan M1 (fix).
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
This gives bench_replay a speedup of ~8x/64x/280x.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
This ensures that stale reads show every write that was answered.
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
This allows a lot of operations to drop mut.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
