| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
After a close vote, one key voting twice passes a rival version at a
history from before the version that followed, the votes cast before
counting with its own, any time until a checkpoint above that version:
nodes that took the version refuse it, but one lagging behind, or
joining from the root, may take it. A node voting by itself proposed a
checkpoint after a version only where its operator set
`--store-propose-versions`; now it does unless told otherwise (1 unless
set; 0 never so). Versions that change more than moving to a checkpoint
are rare, so the cost is a checkpoint each.
A default set, store-propose-versions no longer asks for store-vote-key,
as store-held-for and store-back-after don't.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
federation.md said a client writing once a checkpoint is proposed builds
only on what is above its heads. No client does: a party's client builds
on the heads its node shows, and a node voting by itself on its own,
whatever is proposed. Doing so would let any proposal, which costs one
signature and binds nobody, steer where honest clients build: heads that
never pass, or two proposals beside each other that can't both be
obeyed. A proposer names the heads it holds as it proposes, so what is
written after is above them unless built on an event beside them:
written as it proposed, or on old history; the move refuses that, and
its party's client signs it again.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
`GET /events/counts` says, of each node a party lists, what it said it
speaks and knows (`knows`, by node ID): the wire versions, the oldest and
the newest, and the newest revision of the federation it knows, by
number and ID and as people read it, its number null where this
continuum doesn't know it; null where it said nothing, a build from
before SPEAKS or one the node didn't hear from; the node's own as it is.
So the nodes still to upgrade before a revision passes show, every
party's, where before a party's operators could check only their own.
A party's client, backing a proposal that moves to another revision,
warns of every node of any party not known to know it, a line a party
naming each and what it said: once it passes, those stop at it till
upgraded. `next` warns alike, of the revision it moves to, or of the
later one it suggests. A warning, not a refusal: what a node heard is
no proof, and a node not heard from may well be upgraded.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
A corpus of generated histories, committed, said what this code made of them, right or
not, and held hundreds of kilobytes a revision; generated from code that changes, it
could only be checked, never written again.
Each release that makes a revision is tagged `federation-<F>.<R>`, the federation's
number and the revision's (`federation-1.1`, revision 1 of federation 1).
`scripts/federation-check` writes histories from this tree (`write_histories`, an ignored
test): of each revision it knows and each kind the properties generate, those of a few
seeds from a random one it says, and the attack tests' crafted ones, as each test offered
them where nothing but offers made them. It judges them with `continuum-party judge` here
and with the continuum the latest such tag names (or one given), built in a worktree and
kept between runs, leaves out those whose root the tagged one reads as none, and says
where the two differ. Run before a release, not by `cargo quick`.
The corpus and its test are gone; the test plan and `federation.md` say how a release is
checked.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
A node proposed a checkpoint at the heads of what it held for a while: every event
written in that while, and every one built on an event beside those heads, as one party
writing now and then on something old makes them, was beside them, and refused once the
store moved. A voter whose while was shorter backed none, one whose while was longer
backed it and lost those.
It proposes at the heads its store holds, sorted: what was written before is in their
history, and only what is written while the vote takes its course may be beside them. It
proposes none where they are more than an event may name: under a revision with merge
events, it merges them first; otherwise, what is written after does. `--store-held-for`
keeps the voters' rule and how long a new proposal holds back a checkpoint.
`federation.md`, the guides and the options say so.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Past 64 heads, no event is above them all, and what is written builds on the earliest
64 alone: where nobody writes, heads apart stay apart, and a checkpoint naming them is
never written.
Under a revision with merge events, a node its party lists signs merge events with its
node key where its store holds more heads than an event may name, each on as many as it
may, the earliest taken first, until no more are left (`Holding::merge`). The worker does
so a random wait after the heads first were too many, as it puts again what the document
lacks, so that of the members few do at once.
`federation.md` says so under Nodes, and the guides say what a node writes by itself and
what its log says.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
What builds on many heads merges them, but only an event a party writes: in a quiet
federation, heads beside each other stay so, and past 64 no event is above them all.
Revision 1 (`merge-events`) adds merge events: `{"merge": true}`, signed with a node's
own key, the `key` its Ed25519 key, for the party whose node list names it. One counts for
nothing, and only gathers what it builds on into the history of what builds on it. Under
the base revision, signed by a node its party doesn't list, or by a party's key, it is
refused; the model alike. Generated federations begin at either revision, and their nodes
merge.
A node voting by itself backs no version changing the revision, on any version neither:
its party's operators vote on that one, once their nodes run a continuum that knows it.
`federation.md` lists the revisions, says what a merge event is, and row 71 is tested
(`federation_attack_71`).
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
What every node must agree on was named by the federation's number alone, and a change
that only took more kept it: a party could use it before every node ran it, and nodes
not upgraded refused what followed and parted from the others.
A version now names the revision it follows by its ID (`"revision": "<ID>"`), the
SHA-256 of the JCS list of the features it adds to the base revision federation 1 began
with (`federation::revision`); a version naming none follows the base revision, ECDSA
keys among it, so the versions written so far read as they did. A version names the base
revision only by leaving `revision` out, and a revision no earlier than the one of the
version it changes. A continuum that doesn't know a revision, another build's among
them, can't read a version naming it: it stops at a version event writing one, and says
which revision and what it knows. There are no revisions yet but the base revision.
`federation.md` says how a revision is rolled out, then passed, and what a later
continuum may leave out; the upgrades guide, and the one on a node stopped, alike.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
From the red-team review of the design:
* a party that never votes holds those backing a proposal short by as many votes: they
can neither pass it nor back another, a compromise all others would pass among them;
`federation.md` draws that out, and the promises guide says so;
* a client keeps its own votes and the nays that freed it, not another party's two: where
keys voted twice, whoever saw it keeps the votes naming the party before a checkpoint
leaves them out;
* nothing but agreement keeps a party from using a change that keeps the federation's
number before every node runs it: the next such change names itself in the version, so
that two thirds decide when, and a node not running it stops at that version.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
The votes cast before count at a history from before too: a proposal that died, or a
rival, needs there only the votes it lacked, and after a close vote one key voting twice
is enough, at any time until a checkpoint above the version that followed. Nodes that
took that version refuse it; one lagging behind, or joining from the root, may take it.
`federation.md` said it took the keys of two thirds of the parties.
A node voting by itself proposes a checkpoint once so many versions were written since
its store's (`--store-propose-versions`), as the design lists: with 1, after every
version, so that what keys voting twice grow beside a version ends at a checkpoint above
it. `federation.md` says how little a rival version takes, row 20 too, and the guides the
option.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
A node voting by itself proposes no checkpoint beside a proposal that holds it back, as
the checkpoint would lapse it, and a follower backs none so. A proposal held one back
where someone but its proposer backed it: two parties, or one and a stolen key, kept a
junk proposal alive, and no automatic checkpoint came while it lived, ending it taking
nays from more than a third each round. And one new, its proposer alone backing it, was
lapsed at once by a checkpoint its writes brought about.
A proposal that can still pass holds back now where a third of the parties back it,
someone but its proposer among them; or where it is new, held for less than a while
(`--store-held-for`), and its proposer has none older on the version: so it is heard
before a checkpoint lapses it, once a version for each proposer. Row 67 says so, and a
test of nine parties checks it.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
A node following a party (`--store-follow`) handed it more than the text said:
* it copied the party's `false`: one party's nay freed bound followers, and its `true`
shown to some and its `false` to the rest passed a checkpoint at one history and made it
dead at another. It copies the `true` alone now, on any version too;
* a checkpoint lapses every other proposal: the followed party ended any proposal it
disliked by passing one. A node follows a checkpoint only where no proposal that changes
more can still pass that someone but its proposer backs;
* nothing bounded how often the store moved, so fresh events could be left out of every
checkpoint; and a node just started held nothing for a while, so the voters' rule kept
nothing. A node backs, and proposes, a checkpoint only once its store held its
checkpoint, or ran, for a while (`--store-back-after`, ten minutes unless set), by itself
or by hand.
`federation.md` says what following hands over, and row 58 what a follower on any
version copies; the guides and the example say the rest.
|
| |
|
|
|
|
|
|
|
|
|
| |
No event is above more heads than it may name dependencies (64): a version naming more
for its checkpoint is never written. A node voting by itself proposed one at whatever
heads it held for a while, however many; it now proposes none past 64, as what is
written merges them, and a party's client backs none such (`client::backable`).
A version event names the votes passing a proposal, two thirds of its version's
parties: past 96 parties, more than an event may name. A version names at most 96 now
(`version::PARTIES`).
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Anyone could propose a checkpoint's version again, by an event beside its heads. A node
voting by itself, and a party's client, built a vote on every event proposing what it
voted on, so every vote was beside the heads, and no version event written on them was
taken: the checkpoint could never move the store. And the voters' rule refused it once
that event was held for a while, as the heads left it out.
They vote on the events proposing it that are above the heads it names
(`State::proposing`), which `GET /events/counts` lists as `proposals`; where there are
none, `none` says so, the node casts nothing and the client says why. The voters' rule
leaves proposals and votes aside, as they lapse with the move.
Row 70 is tested (`federation_attack_70`).
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
A key a version names, stolen, could write its party's node list at a history from
before the version that named the party's keys anew: written under the version before,
it counted, as naming keys anew leaves a list's `since` alone, and beside the list the
party wrote under the new version the lower ID won. The thief signed a few, kept the
lowest, and did so again after every list the party wrote.
Of a party's key sets, and of its node lists, those written under the latest version now
come first, then what no other has in its history, then the lowest ID; the model alike.
A list the party writes under the version that replaced the key stays, as `federation.md`
says. Key sets were safe already, by their `since`; one rule serves both.
This changes what counts where two of a party's lists, written under different
versions, are beside each other: nodes read them alike only once all run it.
Row 69 is tested (`federation_attack_69`).
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
A store that stopped can't tell which of what it holds back will be taken, nor which of
what waits: `Store::held_back` gives both, each after what it builds on. A node leaves
out none of it: it puts it again where its document lacks it (`Holding::lacking`), a
remake's snapshot holds it (`Holding::at_base`), and after a switch or a rejoin it puts
again what of it the document it left held (`Holding::left`). A remake while nodes are
stopped no longer drops what was put since the stop, nor the event they stopped at.
Tested: a node keeps what it holds back, what waited among it, through a snapshot, and a
store read anew from that stops there again; and a simulation: a swarm stopped at a
version compacts, a node away meanwhile rejoins and one started again with nothing reads
the new epoch, each stopping there, holding back what came after, until their operators
have them drop that event.
|
| |
|
|
|
|
|
|
|
|
|
|
| |
Of two versions changing the one it reads, a node takes the first it judges, joining or
not, and reports the other: its operator chooses, by dropping and allowing. No rule
holds a joining node back until it has synced with every seed: it learns of the other
members as it syncs, and its seeds may go away.
federation.md drops the joining rule from Changing the policy, and says so in what
isn't promised and in row 63; the guides say what a joining node takes. Row 63 is
tested (`federation_attack_63`): from the root, either side first, the operator
switching sides; and pinned at a later checkpoint, the chain old keys grew never taken.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
A version event writing a proposal passed on the version a node reads, which doesn't
read as a version there (of a later federation, using what a later continuum takes, or
none at all), the node can't judge: `Fold::stops`. The store stops at it: it takes
nothing and refuses nothing from then on, holding back whatever comes, that event first
(`Offered::Stopped`), and keeps none of it across a restart, where it stops there again.
One beside the version the node reads is refused of its own, as any other.
Its operator has it drop that event (`POST /admin/events/drop`), or one it builds on:
it is refused as dropped, or as built on what was, any version event writing that
version with it, kept in the store file; the node judges what it held back and goes on
under the version before. Allowed again, it stops there once more.
The node logs the stop and its end; `GET /events` names the event, why, and how many it
holds back (`stopped`); a put answers `stopped`, written into the document all the same;
a node voting by itself casts nothing while stopped. The test fold's eras tagged `new`
stop a store.
Row 39 is tested (`federation_attack_39`); of rows 9, 10 and 48, a node stops at the
version event.
|
| |
|
|
|
|
|
|
| |
Where a document names a section of another after linking to it
("([`store.md`](store.md), Pinned checkpoints)"), the name links to that
section too, by the id both GitHub and cgit's about filter give its
heading; paragraphs that grew past the width are wrapped anew, links and
code kept on one line each.
|
| |
|
|
|
|
|
|
| |
Where a document names one of its own sections in parentheses ("(What
counts, 5)", "(Joining, below)"), the name links to the section by its
GitHub-style ID; paragraphs that grew past the width are wrapped anew,
links and code kept on one line each. References to another document's
sections stay as they are.
|
| |
|
|
|
|
|
| |
Row 62 names an ECDSA signature's twins, `s` high (the order less `s`)
and `r` or `s` with zeros more in front, each none. Its test offers the
ECDSA vector's write with each (`keys::respelled`) to a store whose root
names that key, and has it take the write as it was signed.
|
| |
|
|
|
|
|
|
| |
A change that only has the federation take what it refused, reading
nothing it took otherwise, as ecdsa-sha2-nistp256 keys, keeps the
federation's number: it is rolled out to every node first, and used
after. A node that doesn't take it refuses what uses it, a version naming
such a key among them.
|
| |
|
|
|
|
|
|
| |
signature low-S
A key is `ssh-ed25519` or `ecdsa-sha2-nistp256`, spelled alike; an ECDSA signature is
over SHA-256, its `s` low and `r` and `s` SSH's shortest mpints, so that it is spelled
one way. The vectors and the agent test, among the harness's.
|
| |
|
|
|
|
| |
What a federation is used for is read and written through a node's API: kartei's
entries are signed in by their parties and read from what counts there, with no stage
of their own.
|
| |
|
|
|
|
|
|
|
| |
The registry holds each kartei namespace in its party's, a place per file. What kartei's
consumers read as one for all (host and user names, the names hosts go by, addresses and
blocks in every net but `internet`) is shared, at places of its own, given by krebs's
rules: a party's by its name, and what kartei gives it beyond them by a rule each in the
first version. What kartei's consumers read is written out from what counts, but for
what names an identifier the current version doesn't give its party.
|
| |
|
|
|
|
|
|
|
|
| |
A node voting by itself, and a party's client, signed a vote on the heads
the node held: where those held what is beside a checkpoint's heads (a
party's writes made while the swarm was parted, say), the vote wasn't above
them, and a version event written on it was refused, so the checkpoint
never moved, though it passed. Both now vote on the events proposing what
they vote on, and a node proposes a checkpoint on the heads it names: so
their votes are above those heads wherever the proposal is.
|
| |
|
|
|
|
| |
The federation's design says that votes a node casts by itself are built: following
parties and proposing checkpoints, backing a checkpoint by the voters' rule, and its
party voting by hand through it.
|
| |
|
|
|
|
|
| |
The federation's design says what of it is built: the store in a swarm's document,
members by the federation and the store at a node's API, as the store's design says; and
a party's client, which signs its party's events, votes by its record, signs again what
a move refused, and names a party that voted twice.
|
| |
|
|
|
| |
store.md, federation.md and federation-tests.md say what of them is built: beside the
store held in memory, the federation as its fold, with its byte formats, and their tests.
|
| |
|
|
|
|
|
|
|
|
| |
The snapshot held the store "as the node reads it at the base". A node that moved to a
later checkpoint since the base forgot what came before it, and reads the store at the
base no more; one that didn't move yet still does: the two built different snapshots
from one base, and the checkpoint the swarm held could leave the document with no node
able to give it back. The snapshot is the store a node reads from the document up to
the base, starting at the document's checkpoint, by its own refusals and drops: every
node holding the base builds the same, whenever it moved.
|
| |
|
|
|
|
|
| |
A party's client votes by what it keeps, never by what a node shows, but nothing said
what one that lost its record does. It can't go by a node: a `true` it sent through one
may be held back there. So it writes its record before it sends a vote, and one that
lost its record casts nothing on the current version.
|
| |
|
|
|
|
| |
No node takes such a history, as each refuses one of the two, but judging, and a model
reading every history anew, are given it, and must say the same of it: it is refused,
as no version is current there.
|
| |
|
|
|
|
|
|
| |
An event doesn't name the version it was written under, and a node starting from a
checkpoint holds none of the kept events' histories. It works out, but `start` has to
assume it, and the text didn't say: a node reads what a checkpoint keeps as written
under its version, as each was written under that one or an earlier and counts there,
a kept proposal is on it, and a later `since` that leaves one out is past it.
|
| |
|
|
|
|
|
|
|
| |
Of two versions changing one, a node still joining takes neither. Continuum's lifecycle
has a node `Joining` only until its first sync, with one peer, its catch-up source: one
node of the adversary's answering first showed it one chain alone, which it took, and
refused the other from everyone else. For this, a node is still joining, from its start
or a rejoin, until it has synced with every seed its operator named; and what isn't
promised is a node joining whose seeds are all the adversary's.
|
| |
|
|
|
|
| |
Once a checkpoint passed, a node joins from a pinned one; but a node started from the
root can't tell that one passed, as a chain old keys grew need name none. It is its
operator's to know, not the node's to refuse.
|
| |
|
|
|
|
|
|
| |
`keys`, `nodes` and `proposal` events had no shape written, a vote's was written as if
it were the event, and the fold had no name, though the genesis change is built byte for
byte from the policy naming it. Each field an event may have is written with what it
holds, a version's hash as an ID is, and the fold is named `federation`:
`{"version": 2, "store": "federation"}`.
|
| |
|
|
|
|
|
|
|
|
| |
An event's `key` was spelled out; the keys a version and a key set name, and node IDs,
weren't, and whether a version is one hung on it: a key's base64 not canonical, a
comment, 32 bytes that are no Ed25519 key, a key twice in a set, a node ID in capitals
or twice. Keys are spelled as `key` is, with no comment, their 32 bytes an Ed25519
public key; node IDs as continuum spells them, `continuum:` and 64 lower-case hex
digits; and none of either twice in a list. The example version loses its keys'
comments.
|
| |
|
|
|
|
|
|
| |
The design named the blob's hash and key, and said a signature is checked as continuum
checks a user's, which takes `sha256` beside `sha512`, and any reserved field. Only the
signer can vary that field, as it is signed, but one builder asked it empty and another
didn't. It is empty, as `ssh-keygen` writes it, and the hash `sha512`, beyond what
continuum's check asks.
|
| |
|
|
|
|
| |
256 KiB hold arrays some hundred thousand deep. A JSON reader stops at a depth of its
own, or none, or falls over: two builders' nodes parted on such an event. An event
nests objects and arrays up to 32 deep, its own object the first.
|
| |
|
|
|
|
|
|
| |
Integers were asked of values alone. A proposal holding `1.5` or `1e21` is good JCS, so
to judge it a builder needed all of ECMAScript's number printing, for this alone, and
one reading integers only took it for no event. Numbers anywhere in an event are
integers from −2⁵³ to 2⁵³, and strings anywhere are whole Unicode characters: floats,
and the hardest part of JCS with them, go.
|
| |
|
|
|
|
|
|
|
|
|
| |
The sizes federation 1 allows listed a party's keys "in its version or its set" among
what makes an event none, while a proposal is taken whatever version it holds: a
proposal giving a party 17 keys was no event to one builder, and a proposal that is no
version to another, who parted at the first event with it in its history. The list now
says what of an event makes it none (its size and dependencies, a key set's keys, a node
list's nodes, a write's path, its values), and a version is none where it gives a party
more keys than a key set may hold, more nodes than a node list, or a rule a longer path
than a write's.
|
| |
|
|
|
|
|
|
|
|
| |
A vote on a version naming a checkpoint was taken only where it was above the heads. It
follows from the version event's being above them: the version event is taken only
there, and with it every vote it is written on, as everything in the history of an
event above some heads that isn't in theirs is above them. Two rules are left: the
proposal is a version only where it is above the heads, and a version event writing it
is taken only where it is above them. A vote cast beside them binds its party as any
yes does, until the move refuses it; an honest client builds only on what is above them.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
`true`
By default a node followed a party's `false` on any proposal. Where that party and its
followers are more than a third of the parties, one key, the one kept on a node, ended
every proposal, the one naming that key anew among them. By default a node now follows
only on a version that only moves to a checkpoint, both ways, and following on any
version, `false` among it, is its operator's choice.
The follower's `false` was also how a proposal nobody else backs stopped holding back
automatic checkpoints. Instead, a node proposing a checkpoint goes ahead beside a
proposal it wouldn't back that nobody backs but its proposer; it still waits for any it
would back, or that others back.
|
| |
|
|
|
|
|
| |
A node following a party backed a proposal that party proposed, but a proposal binds its
proposer to nothing: its followers were bound by their yes, and the party it followed
was free to back another. A node follows a `true` only, and one proposing a checkpoint
backs it, so that the nodes following its party back it too.
|
| |
|
|
|
|
|
|
| |
A party votes from one client, and may have one node vote by itself, but nothing put the
two together: a node following a party's `false` on a proposal, and its operator backing
the same proposal from a client, is a party voting twice, honestly. So such a party
votes by hand only through that node, which keeps the party's one record of its votes
and refuses a vote the record forbids.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
for a while
A version that only moves the store on refuses what isn't above its heads, which its
proposer chooses: heads built on little, just after the latest version event, would
take back what was written since, a key set dropping a stolen key among it, and the
stolen key would count again; a node following a `true` on checkpoints, or a voter
checking only the version's fields, would back it unseen.
So a party's client and a node voting by itself back a checkpoint only where everything
they held for a while before it was proposed is among its heads, in their history, or
above them, and a node shows what the heads lack, as for a fork. It is the voter's rule,
by its own clock; nothing more is judged. A row for the attack table, and the property
of votes a node casts by itself, say so.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
A merge, a fork at heads of several stores, named every party's keys anew, so that a
party wrote its key set again before anything a key of its set signed counted there;
and kept each side's values side by side, so that a value a party replaced on one side
came back where the other side's write had the lower ID. Letting the other side's
parties in, by a version or a fork naming them, who sign again what they keep as a
party's client signs again what a move refused, does the same with nothing new.
So a fork is at heads of one store: the fold makes its root from that store's state,
and the store from its own heads (`Store::fork`). Merges' rows go, those after them
move up, and Roads not taken says why, with merging in place.
|
| |
|
|
|
| |
store.md, federation.md and federation-tests.md said none of them was built; the store
held in memory is, with a fold made for its tests and their properties.
|
| |
|
|
|
|
|
| |
A node starting from a checkpoint its operator pins holds neither the checkpoint before
it nor the events between, so it can't check that the version naming its checkpoint
names heads after the one current at that version, at which that one is current too. It
takes that on its operator's word, and so takes the version, as every node did.
|
| |
|
|
|
|
|
|
| |
A version names checkpoint heads after the current checkpoint, and now heads at which
that checkpoint is current too: the version naming it is among them or in their history.
Without that, the version naming the current checkpoint could be after the next one's
heads: every node that moved there holds it, and one starting from that checkpoint,
which can't check heads before its own, couldn't judge it as they did.
|