aboutsummaryrefslogtreecommitdiffstats
Commit message (Collapse)AuthorAgeFilesLines
* federation: brought along, and switch recordstv3 days1-86/+88
| | | | | | | | | | | "Carried over" is continuum's word for replay: a node's writes written again by path, the last writer winning. The federated swarm does something else after a remake: envelopes written again as they were signed, and what each node held of them recorded. Envelopes are now "brought along", and a node's record of what it held when it switched is its "switch record" (`_continuum/switched/<node ID>`). The Remakes section says that continuum's replay doesn't run under the federated policy.
* federation: what it asks of the document, and where Automerge sets limitstv3 days1-113/+170
| | | | | | | | | | | | | | | | | The design leaned on Automerge in places it didn't say: writers that cost memory, one change per writer and number, changes Automerge applies and shouldn't, the order of values written at once, maps made at one key, history held in memory, bytes that versions must agree on. A section near the start says what the design asks of the document, and nothing else of it is read: changes named by their hashes, each judged at its history; the document at a set of changes, the same on every node; a writer for each change; history kept as long as the store keeps it, and remakes. A section at the end says, for each of Automerge's seven limits, what Automerge does, what the design does about it, and what would go without it. Every rule that is there only for one of them points to it in place. "Writer" is the design's word for the identity a node writes under; "actor" stays with Automerge.
* federation: the words this round added, for the glossarytv3 days1-3/+5
| | | | | | | The words the glossary gets with the federated case: final, of a decision too; stopped counting for good; the voters before; and the alive peers and their acks, as the design names what continuum calls the electorate and its votes.
* federation: leftovers of rules since made generaltv3 days1-14/+14
| | | | | | | | | "By any node" under Decisions said for votes what is now every statement's rule; it says so in a line. And the voters no longer remake: the row of two thirds turning says what they still can. A matter's examples name who votes and who the roots are; and the list of those who take part goes by who voted or signed anything that stands, since an epoch no longer spans a term.
* federation: what to watch, and when to acttv3 days1-2/+54
| | | | | | | | The design showed much and said nowhere who looks at what, nor when to act. A section says it for each who acts, a node's operator, a user, a voter and a root: each signal, where it shows (an endpoint, and a metric beside it), the threshold the design recommends, and what to do. What a node put off shows (`GET /admin/put-off`), beside taking it.
* federation: what it coststv3 days1-6/+52
| | | | | | | | | | | What every node pays was scattered over the text: what a statement, a stamp, a cut's list and a carry record take, a snapshot a remake, a signature's check. A section of its own says it per thing (a change, a writer, a statement, a value, a node, a user at its limits, a claim, a cut, a root's node, a remake, a signature), in memory, on disk, on the wire and in time; what bounds the changes and the writers, and so an epoch's heap (some 42 MB at the remake numbers recommended); what nothing bounds; what was measured; what is assumed cheap.
* federation: a user's claims of one name, counted once, released togethertv3 days1-166/+171
| | | | | | | | | | | | "A user's own claims of one name are one, ordered against others by the earliest" left three things open. They are ordered by the first of them that stands, by the numbers they state and then the digest, so a duplicate committed after a release doesn't carry the released one's witness. They count once toward the kind's limit, so a node that held duplicates back can't commit them to push the user's later claims past it. And a release gives up every claim of that name of its user but those signed in sight of it, so a user gives a name up with one release, though a node may still hold a duplicate it handed out.
* federation: the payloads hold what the rules readtv3 days1-21/+29
| | | | | | | | | | | | | | | | | | | | | Six places where a rule read a field the payloads didn't have: * a value's envelope, with its `prev`; and since a value says nothing of what its signer saw, where it was signed is the change that carries it; * `unvouch` of a user the root change holds without a creator, which has no `create` or `vouch` to name: `for`, its ID; * `release` of a name a decision gave or made over, which has no claim of its holder's to name: `claimKind` and `name`; * `firstKeys` holds its cut's `keeps`, as every cut does; * the matter `name` names the kind beside the name; * a member entry says which open proposals its bounds would refuse (`refuses`). `key` named three things in one payload; the signing key is now `signingKey`, the key a `key` statement adds `added`, the one a `retire` retires `retired`. And `user`, the signer's name beside its ID, goes: nothing compared the two.
* federation: catching up has a number of its owntv3 days1-5/+6
| | | | | | | | | | A node that comes back writes nothing before it caught up with the peers it keeps, or `witness` passed: one number meaning two things. A peer that delays what the node lacks for a minute had a root's node, back from a day off, write blind. `catchUp`, recommended 10 minutes, lets a swarm choose how long such a node waits. Past 32 heads, the STATUS names them by a digest, so the node goes by the sync that follows it.
* federation: what a party's budget countstv3 days1-26/+35
| | | | | | | | | | | | | | | | | A node takes a party's changes no faster than a budget, and the design never said what the budget counts nor recommended one. It counts changes, their bytes and their operations an hour, recommended 600, 4 MiB and 100 000 a party: a change of a few bytes that holds a million operations draws on it too, and a change larger than what is left of the hour is taken all the same, its party's next changes waiting until it is paid. A node's changes draw on the party of the user its writers count for, whatever they carry. A root's node wrote once more for every claim that reached it, so one party's claims and releases had every root's node write without end. It now writes once a `witness` at most, for all claims that reached it meanwhile: 1 440 changes a day with a minute, within its party's budget.
* federation: `limits.writers` recommended at 64tv3 days1-6/+9
| | | | | | | | | | Every writer costs every node 4 bytes for every change of the epoch, those before it started too. At a year of a hundred nodes' daily stamps, one user that starts 1 024 writers costs every node some 143 MB; 64 cost some 9 MB there, and 2.5 MB at the `remakeChanges` an epoch now reaches. 64 writers are a node restarting weekly for more than a year; a swarm whose epochs last longer than that (few nodes, writing little) sets it higher, as the text now says.
* federation: a key's own retirement takes back nothing seentv3 days1-6/+9
| | | | | | | | | | | | A key retires itself only from where it stands, its cut holding all it signed that the retiring change's history holds. But whoever commits the retirement chooses that history: on dependencies from long ago a stolen device key retired itself from there, and what its user had signed with it since was outside the cut. A key's own retirement now takes back nothing that a node of its user or of a root had in its history without it, as far as an unvouch reaches. A key above still cuts a stolen key off from before the theft.
* federation: the voters' cut takes back a root's creationstv3 days1-20/+26
| | | | | | | | | | | | | | | | A root's creation outlasts a cut of that root's key where another root's node saw it, so that a root that turns can't take back settled or final names. A thief of a root's key kept the same way every moved-in user it gave first keys to, past the root's cut and past the voters' too, to be undone one decision a user. A cut a user sets (the root itself, or a thief of its key) still leaves what another root's node saw; the voters' cut, by first keys they give a root, takes back what the cut-off key created outside it, whoever saw it: two thirds of the voters undo a stolen root key in one decision, and no root does alone. With one root nobody else saw anything: its cut takes back a name only its creation holds, final or not, as the text now says beside the one voter.
* federation: a claim in sight past an unvouch or a cut, by the roots' nodestv3 days1-96/+100
| | | | | | | | | | | | | | | | | A claim signed in sight of another that was down there only by an unvouch or a cut is ordered against it by lateness, over all the changes. The text counted the nodes of the other's side, the other claimant's own among them: a claim hidden, cut off and brought back then contested one made rightly in sight of it, and an unvoucher's own stamp, made weeks later, brought a claim back over one made in sight of the unvouch. For these pairs the roots' nodes alone are the witnesses: the claimant saw the other, and the unvouch or the cut, and said so, so neither its own old stamps nor the other claimant's nodes count. Two claims and Resolving say so too, the leftovers of the earlier reading go from the "else" under Claims, and the registry shows such a claim as held while the unvouch or the cut stands.
* federation: nothing is left out with the node that carried ittv3 days1-182/+183
| | | | | | | | | | | | | | Any node commits any user's statement, yet a statement still counted only where its node committed it within that node's cut: a party that stopped its node with a cut from before took back what the node had committed for anyone, a settled claim among them, and an accomplice had the name. A node's cut now takes back what the node witnessed and the writers it started, never what it carried. With it go the leftovers of the rule that only a user's node carries its statements: the attack row and its test, the trust table, the API. An out voter's node commits its carry record beside its records, and peers take its changes for both.
* federation: a moved-in user's first keys come from the root named for ittv3 days1-54/+64
| | | | | | | | | | | | | | | | | | The root change names, for each user it holds without keys, the root that gives them; no other root's `create` gives keys. Two roots' creations no longer compete, so a contest arises only where that root signs two: they are ordered as claims are, with the other roots' nodes alone as witnesses, so a twin on old dependencies is late for good once another root's node saw the first. Which creation gives the keys is read over all the changes, as which claim holds a name is, and no longer where each statement was signed: read there, a twin was the only creation at whatever place its maker crafted, and its keys signed as the user. While two contest, nothing signed with the keys of either counts, until the voters give first keys; what the keys they give signed counts then. A decision on the roots that leaves a root out says which root gives the keys it was named to give.
* federation: remakes are continuum'stv3 days1-465/+375
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | A federated swarm remakes as every swarm does: a node proposes and coordinates, the alive peers ack, every node builds the next epoch itself. The voters no longer decide remakes, so the base check, the votes naming a base, `baseAge`, `remakeAge`, the heal, and the admin actions `choose`, `follow` and `back` go. What the federated policy adds: * compaction only, and no sooner than the epoch's history holds `remakeChanges` changes (recommended 10 000) or its changes started `remakeWriters` writers (1 000): every node proposes and acks by these numbers, and follows only such a remake; * any base serves: the snapshot holds the registry at the base with whatever something signed later could still change (envelopes with their signatures, what was given up and hasn't stopped counting for good, what the order among claims rests on, how far each length of time has run), and every node carries over what it holds beyond the base with what it held of it; * envelopes of the epoch before come from any node, as within an epoch; a claim carried over and one the snapshot holds are ordered as two claims out of each other's sight, not by the base; * lengths of time go on across a remake, each node's carry record saying what it saw broken after the base; * of two remakes of one epoch, every node goes to the one with the lower ID. A remake made a cut reach no further back than its snapshot. With remakes as frequent as the document grows, that would leave a stolen key's work standing after an hour. So what was given up, replaced, left out by a cut or past its limit, and was seen so for `final`, has stopped counting for good, remake or not; that is what a remake leaves out.
* federation: who votes and who the roots are, by decisionstv3 days1-172/+185
| | | | | | | | | | | | | | A remake no longer names the voters and the roots: each is a matter the voters decide, as a number is, and holds until another decision of it. Whether a user is a root is read where each thing it did was done, so a decision changes nothing done before it. What the votes of before could still do once others vote, a remake used to put out of reach. Two rules do it now. A decision of voters that a decision on who votes replaced counts only where that decision's proposal was signed in sight of it decided. And a decision seen holding for `final` is final: no decision of its matter whose proposal wasn't signed in sight of it holds instead, whatever its number.
* federation: what each payload holdstv3 days1-0/+59
| | | | | | | | | | | The design named fields in passing and listed none, though stage 2 settles the payloads first. A section lists the fields each rule reads: every envelope's (swarm, epoch, path, kind, signer, what it saw, its number, the key and signature), each statement kind's, a proposal's matter and what it holds (a remake's switch delay among them), a vote's; a fork's vote, which counts toward no matter; and what a node writes plain: a record of a proposal, a carry record, a member entry with the user its writers count for and the digest it computes.
* federation: operator actions, writers past cuts, caught up, fork bases, relaystv3 days1-54/+87
| | | | | | | | | | | | | | | | | | | | | | | | | | - What an operator says, listed with each action named where it is used: rejoin, choose (two remakes decided, or a walk through the archives that ends elsewhere), follow (a node that waits past `remakeAge`), back (refusing the remake), fork, refusing and bounds, dropping what the others refuse, accepting a check of itself that differed, taking what was put off. Tokens from the nodes `--token-nodes` names. - Where a cut takes a thief's node's writers off a user's count, below the limit again, every node takes what it put off for that user, nobody acting. A writer counts for the user it was counted for until such a cut. - Caught up: the heads each peer kept had when the node reached it, as its first STATUS says, or `witness` passed; a peer that keeps sending keeps it silent no longer. - A fork's base: rule 4 takes its proposer's node's record, without votes; it refused it unless two thirds voted for the fork. - Relays: each node's configuration in a federated swarm; nobody may write `_continuum/relays` plain. - In the open: `GET /registry/users/<name>` shows which root's node keeps a user in past an unvouch or a cut, and which roots' creations contest a moved-in user's first keys. - A carry record names what counts through a cut, as through an unvouch. Two rows move to the stage of policies changing: a limit lowered and raised again, a stamp between.
* federation: going back refuses, every node takes a base up, the heal's fetchtv3 days1-99/+119
| | | | | | | | | | | | | | | | | | | | | | | | | | | - Going back by an operator's word refuses the remake: the heal doesn't move such a node again. A node sent back followed again once the count crossed a third, and what its users had signed again came along beside what they signed in the new epoch: two users of one name. Meanwhile neither epoch decides what takes nodes of more than two thirds of the voters; a node moves on by word. - Every node takes a base up, and counts `remakeAge` from then; a node of no voter notes when, and builds the snapshot only once the remake is decided. Taking a base up chooses nothing. A node of no voter had no clock, and a remake shown long after its time would have taken it. - Every node that switches commits a carry record, out voters' nodes too: the heal's third is of all the voters. - A node that waits asks a peer in the new epoch for its carry records once, when they connect, and is pushed each one the peer takes after: nothing is polled. Stage 14 builds it. - A carry record speaks for the epoch it leaves: its roots witness, a third of its voters say what was seen for long. - The proposal names its switch delay; one naming no root is no proposal; a moved-in user's contested creations go along in the snapshot. Who undoes an unvouch's witness, and how far back. The malicious choose when the count crosses, by their own records: said. Attacks: a late remake against nodes of no voter; back and forth by a split that stays.
* federation: claims past cuts by lateness, roots past their cuts, any committertv3 days1-202/+223
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | - A claim made in sight of another that didn't stand there only by an unvouch or a cut is ordered as claims out of each other's sight, over all the changes. Judged within the history its signer chose, a creator that cut its key off from before the creation, on old dependencies, then claimed the user's settled names there, found no witness: the user's claim gave way for good, no root taking part. The roots' nodes that had the user's claim count first now make the new one late; a root's node witnessing after the fact makes both late, a contest. The exception for a root's witness of an unvouch goes. - A moved-in user's first keys are read where each statement was signed, the roots' creations ordered first, by the roots' nodes alone (a user's nodes carry for it by the keys these decide). Read over all the changes, a root's twin on old dependencies took back everything the user signed; now it freezes the user until the voters say, and takes nothing back. - A root's creation another root's node had seen before the cut stays the claim of its user's name, and a moved-in user's first keys. A root that turned took its creations' names, final ones too, where only the roots create users; now only what the other roots hadn't seen. A thief of a root's key keeps what it created in their sight, for the voters. - Any node commits any user's statement but a `node` statement: which node commits one, or whose, says nothing of it. A user's own claims of one name are one, whatever node they came by. Attacks: a creator's claim in sight past its cut; a root cutting its creations; a root's twin taking back what the user signed; a claim through any node. Rows 7, 9 and the new twin row move to the stage of claims out of sight.
* federation: no root release, roots witness past cuts, an unsteered healtv4 days1-240/+261
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | - No root releases its `create` for a user the root change holds. Allowed while it contests another root's, a root made its own contested by a twin on old dependencies, or signed the release on old dependencies once it was final, then gave the user keys of its own: nothing the user signed counted any more, final names included. Two roots' creates contest until the voters give first keys. - A root's node witnesses past a cut, as against an unvouch. With only the signer's nodes, a creator that turned cut the creation's key off from before it, its nodes silent, and nothing its user signed counted: an accomplice took the user's settled names. A root's node that witnesses for a thief's sock keeps it in, in the open; said under what is not promised. A carry record witnesses for a vouch or a creation cut off as for an unvouch. - The heal of a split by `remakeAge` counts a node for one voter, of the voters of the epoch that decided the remake, each carry record as it stood where written; a node that waits fetches them from peers in the new epoch. A node that followed goes back only by its operator's word, and the heal moves only nodes that wait by `remakeAge`, not one whose operator refused the remake. The malicious could steer whether a split healed forward or back, and send the honest back and forth, each time at the cost of what their users wrote; they choose when a split starts, and no more. Argued under why it holds. - Said: own-name claims where only the roots create users, and under a rule a decision changes; a `remakeAge` of nothing ends remaking; the cut's price with the same witnesses; which member entry a writer counts by; screening's generated changes among what is assumed. Attacks: a creator's backdated cut with its nodes silent; a root's twin `create`; a heal steered by carry records held back or taken back. Rows 8, 96, 105 and 106 and row 129's test name mended.
* federation: a split by a remake past remakeAge shows, and healstv4 days1-27/+49
| | | | | | | | | | | | | | | | | | | A remake shown when some honest nodes are past `remakeAge` and some aren't split them for good: those within followed, those past stayed, their remake votes spent. It shows on both sides now (those that wait hold the decision and see peers in the new epoch; those that followed see peers in the old one), and heals by the new epoch's carry records, which every node reads alike and which only grow. A node past `remakeAge` waits in the epoch instead of staying for good, and follows once they count for more than a third of the voters: a node of an honest voter got the decision in time, and the malicious alone count for less. While they count for no more, a node that followed goes back `remakeAge` later, dropping the new epoch as a node comes back from changes the others refuse, its users' writes there quarantined to be signed again; should they cross a third later, it follows again. Every honest node ends in one epoch, and a remake shown long after its time is followed by the malicious alone.
* federation: how late a remake a node follows, a number of the policytv4 days1-56/+63
| | | | | | | | | | | | | | | | | How late a remake its node follows was each operator's bound, by its own clock: operators set different bounds, and nodes take a base up at different times, so a remake shown when some honest nodes are past their bound and some aren't splits them between two epochs, those that stayed with their remake votes spent. It is a number of the policy now, `remakeAge` (30 days recommended): how long after taking its base up a node still follows a remake. The bounds no longer differ by operator; that nodes take a base up and get the decision at different times remains, and with it the split, said under what is not promised and in an attack row. The bound isn't read from the changes, and a remake past it doesn't free its votes: a deciding change built on old dependencies always looks in time, so the remake could still be decided beside the one its voters then voted for.
* federation: roots keep their creates, writers by entry, cuts by the signertv4 days1-210/+233
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | - A root releases its `create` for a user the root change holds only while it contests another root's. Released at any time, it took the user: none of its keys stayed its own, nothing it signed counted, final names included, and the root gave it keys of its own. - A node's writers count for the user its member entry names, its own word, carried over a remake with the entry; the lowest digest of a `node` statement only where it names none. By digest alone, after a remake, any user the node carries for could make itself the node's, by signing until its digest was the lowest, and have the node put off as a whole. - A user's claim of its own name is valid only where a rule of the kind `user` would let the user create that name: a name kept for one creator's creations goes with that creator's cut, as the attacks said. - Past a cut, only the signer's nodes witness what a vouch or a creation carried, not a root's: a root's node could keep a thief's sock in for good, committing with a fresh writer on heads without the retirement. An accomplice's vouch keeps a sock in; said under what is not promised. - Clients offer a vote for one of the remakes under way that name the same voters and roots (most votes, then the lowest digest), and none for a proposal whose coordinator named two bases; that one voter can still split honest voters signing within the same minutes is said. - A snapshot built for every remake's base is among what an epoch costs. - What screening refuses: tables and marks, a deletion named as replaced; the test that generates changes. Leftovers: decided, for a remake, in Decisions; rows 60 and 93 to the stage of unvouching. Attacks: a root releasing its `create` after its user signed, a user making itself a node's across a remake, a twin remake, a root's witness past a cut; row 130 with the three kinds screening refuses now and their tests.
* federation: every base taken up, writers for one user, cuts as unvouchestv4 days1-292/+358
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Remakes. A node took up the first base it saw and no other, beside votes that name their base and are never released: two remakes proposed at once, or a base handed first to nodes back from away, split the nodes, and a voter whose nodes took one base up could vote for no other, ending the epoch's remakes with nobody malicious. A node now takes up every base that passes, one record a proposal; records count toward taking a base up however old without votes, and a remake's record is judged without them (as written, a record needed votes in its history that come only after it). Clients offer a vote only for a base whose records count for more than two thirds of the voters, showing the other remakes under way. The votes alone keep two bases from both being decided. What votes never released cost is said under what is not promised: voters who split end the epoch's remakes, and whoever holds a hanging remake's last record chooses its day; an operator bounds how late a remake its node follows, and refuses a later one by staying. A remake's vote runs while the epoch lasts, and the voters don't end it by a later proposal; said where every vote was said to. Writers. A node carrying for several users counted its writers for each, so one user past its limit put off every node that carries for it, for all their users. A node's writers count for one user now: the first that named it, as its own changes show. Names. - A creation or vouch cut off stops counting as its signer's unvouch would, from where the cut was set: a user only its creator vouched for is out from there, its names waiting, final ones too, until it is gone or a vouch comes. Before, everything it signed stopped counting at once. - A root releases its own `create` for a user the root change holds: two roots answering at once contest, and nothing ended it. - A `create` is refused where a user's claim of its own name holds the name, and a user has one such claim at a time. - A root's witness outside a claim's history makes two claims contest only where it had the other count through its own witness of an unvouch, as meant: a limit lowered and raised again leaves sight to decide. - What comes back over a later claim is said for a holder's own release, limit or cut too. Attacks: two remakes proposed at once, a remake held by its last record, a node carrying for a user past its writers, a creator cutting off a user only it vouched for, a crafted first change of a writer in a batch, a user's own name claimed again; row 124 with what transactions never write and the tests that show it. A limit lowered and raised again moves to the stage of decisions, a witness after an unvouch to unvouching. What was checked: the committer is named from a change's own history.
* federation: votes name a remake's base, a user's own name, writers per usertv4 days1-154/+206
| | | | | | | | | | | | | | | | | | | | | | | | | | | | The third reading against the text and the code: * A remake's base: the coordinator names it, a voter's vote names it, once an epoch, never released, and a record counts for a voter only where its vote names the record's base. A coordinator handing two bases to two nodes of each voter gets one decided or none. A base whose records count for a third of the voters every node takes up, however old: a partition at the base no longer ends the epoch's remakes. * Whether a claim stood where a later one was signed is read with the witnesses in that history; one made after brings nothing back, and a root's makes the two contest. A claim pushed past its limit by statements a cut leaves out comes back, as one released by a stolen key does. * A user claims its own name with its own key, which holds the name where its creator's key is cut off from before its creation. * A decision on whose a name is makes every other claim give way, two final ones too. * A root's creations giving keys to a user the root change holds are ordered as claims: one root coming after another gives none. * Writers are counted per user, as settled; a twin switches a node's writer only signed with its key. A node keeps when it received what. * Leftovers: an out voter's records taken from its nodes, the lists of a cut, the encoding among what can't change, a change's operations among what isn't bounded, stages, words. 129 attacks.
* federation: a root is a user, an epoch's first change its root changetv5 days1-38/+39
| | | | | | "Root" meant both a user and an epoch's first change. The first is now always the root change. The budget's test goes with the budget (stage 8), and a crafted change's tests, which exist, with stage 1.
* federation: what a cut brings back, what `prev` namestv5 days1-114/+132
| | | | | | | | | | | | | | | | | | The questions the design left: * A cut that takes back an unvouch brings its user in again. Its claims from before stood all along; what it signed while out comes to count later than it was signed, and gives way to what was claimed in sight of it. Gone stays gone. * A creation taken back by a cut is taken back as a vouch and as the claim of the user's name, not as the user: a vouch names a user by its ID, so with another vouch the user stays in, keys and claims, without a name. * `prev` names the value that shows where the envelope is signed; where several count there, any one of them, and the write replaces them all. * A carry record needs telling from no other change: it witnesses by what it names. A remake's proposal isn't carried over.
* federation: carry records, refusing within a window, digests by two thirdstv5 days1-161/+228
| | | | | | | | | | | | | | | | | | | | | | | | | | | | What was left open after the second reading: * After a remake every node first commits a carry record: the digests of what it holds beyond the base and what it held of it. It witnesses as the node's changes did. Then each envelope is written again once, by whoever finds the epoch lacking it, and only where a carry record names it. So a late remake doesn't bring every envelope once per node, and nobody hands in later what no node held at the switch. * What is carried over comes after what the snapshot holds and before what the new epoch signs: a number nobody could check orders only its signer's statements, and a carried proposal outranks none of the new epoch. A value of the new epoch has the last word over one carried over. * No node takes up a proposal that says it saw less than the node held a window before, so whoever proposes can't make refusing cost a whole epoch. * A node stops writing only where its digest differs from the nodes of more than two thirds of the other voters: fewer than a third stamping nonsense stop nobody, with three voters too. * A node carrying for a third of the voters, keeping its records back, is said, with what the voters do about it; a voter without a node names carriers of two parties. * Writers of a node stopped with a cut count for nobody. A limit lowered for a day gives names away for good, said.
* federation: what an operator does on its nodes is no part of ittv5 days1-90/+79
| | | | | | | | | | | | | The design asks that an honest federation bears whatever the malicious do, and rules nothing about a party's own nodes, which nobody could enforce. So: * Bans go: `ban`, `--follow` and their limit. Leaving a user out when reading is each reader's own, and a side note. * Whom a node lets in and what it lets them do is its operator's. Keys in the registry give nothing on a node by themselves. * A remake that leaves a root out says which root has each name a rule gives a root, so none is without a holder.
* federation: one base an epoch, sight decides alone, lists of voterstv5 days1-358/+477
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | The design after a second reading against itself and the code: * An epoch has one remake: a voter's node records one base for its epoch and no other, and a later proposal is taken up only for that base, those voters and roots. No node that switched moves again by itself. The proposal lists the voters by their IDs and names its coordinator; nothing after it is signed changes the list. * A voter votes for its epoch, in, out or gone, and the nodes of one that is out commit their records and nothing else. A node takes up only a higher number than it took up. A record is read where it was written. * Of two claims, one signed in sight of the other: sight decides alone, asking whether the other stood where the claim was signed. Late is asked only where neither saw the other, and a claim late against one that isn't late against it is late for good. A final claim gives way to none but another final one. An earlier claim comes back over a later one only where a root's node had seen it hold the name before it went down. * What was given up or is left out by a cut comes before nothing, so a stolen key's statements stand in nobody's way once it is cut off. * A retirement lists what its cut keeps, so the cut holds after a remake. After a remake a node states what it held of what it carries over (which claim late against which, what counts through an unvouch, which value replaced which), where an order of changes said it; of several values at one place, one that another replaces is left out. The snapshot keeps the proposals that are open. * Giving a user up puts it out, and it is gone after `gone`: a cut takes it back until then. First keys for a user the root change holds are given once. * Writers are counted by a party's budget and put off past `limits.writers`, never refused. A party, for a budget, is a user with the users only it leads to. * A node that follows several users takes a ban where most of them state it. Between two stamps is read from their histories. The relays stay an operator's list. * What is asked of a change before Automerge sees it, as the code has it. 121 attacks.
* federation: a thousand writers a user, and what the limit doestv5 days1-2/+6
| | | | | `limits.writers` is recommended at 1024. A user whose nodes started that many starts no more until a remake; how many they started shows.
* federation: changes Automerge wouldn't write are refused todaytv5 days1-12/+13
| | | | | Screening refuses them in every swarm, so the design builds on it and no stage adds it.
* federation: why not an actor for every changetv5 days1-0/+18
| | | | | | | | | | | | | | An actor that is the hash of its change with the actor left out makes a fork of an actor two changes that every node holds: it can be made with Automerge 0.12's API and is checked with one hash. But Automerge keeps a slot per actor in every clock it caches and moves its indices for every actor that comes, so with an actor per change heap and time grow with the square of the changes: 32 to 257 MB and 2 to 20 ms per change applied at 10^4 changes, 2.7 GB and 0.16 s at 10^5 (measured with Automerge alone). A hundred nodes stamping daily write 10^4 changes in a hundred days. So a fork of an actor stays proven and named, not prevented.
* federation: a limit on the writers a user's nodes starttv5 days1-30/+52
| | | | | | | | | | | | | | | | A writer (an Automerge actor) costs every node heap in proportion to the document's history: about 4 bytes per change, a slot in each clock Automerge 0.12 caches. A document of 10^5 changes takes 42 MB with 10 writers and 435 MB with 1000 (measured with Automerge alone). Screening counts no actors, so a party starting a writer with every change adds gigabytes to every node. `limits.writers` (256 recommended) bounds what a user's nodes start in an epoch, judged as a change's history shows, and a federated node keeps its writer over a clean stop. Out of each other's sight such writers are valid and held: the nodes that made them are refused as peers, a party's budget bounds how fast they came, a remake leaves them out. Not promised: a bound on what an epoch's document costs.
* federation: only changes Automerge itself writestv5 days1-18/+41
| | | | | | | | | | | | | | | | Automerge 0.12.0 applies changes no honest node writes: one that lists an actor none of its operations names, one whose first operation is numbered below its dependencies' or at 2^32 - 1. It then gives back another change than the one it was given, saves a document that doesn't load, or panics (measured with Automerge alone). A party that crafts one stops the honest nodes, where a forked actor only parts them. So a change is refused before Automerge sees it unless its bytes are the ones Automerge encodes for it, every other actor it names has changes in its history, and its first operation is numbered one after the highest of its dependencies. An attack with its test, in the first stage.
* federation: several roots, final names, every voter countstv5 days1-820/+1168
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | The design as it stands after being read against itself and the code: * A statement says where it was signed: its payload names the changes its signer saw. "In sight of", the order past a limit and a proposal's number are read from that, so carrying a statement a second time moves nothing. A statement that is counted states its number. * A node carries for a user once it committed the statement naming it itself. Tokens and logins by key are taken as a node's operator says. * Lines say who is in and nothing else. A claim's witnesses are its user's nodes and the roots'. * The roots are several, named by an epoch's root change like the voters: by the genesis change, then by each remake. What a root signed for another is the seats' once its signer holds none, and the roots take that back only together. * A claim seen holding its name for `final` by a third of the voters is final: no rival holds or contests the name after, whatever a root says and whatever is cut off. It takes no root. * A claim made in sight of one that didn't stand doesn't give way to it when it comes back (its user vouched for again, a limit raised); only a release or a cut that a cut takes back does that, until the later claim is final. Late is witnessed by a change that saw the other stand while the claim didn't. * A cut never leaves out its own statement, and takes back what the key it cuts retired, stopped and unvouched. Giving a user up takes a first key. * Every voter counts toward a decision: more than two thirds of all. Nobody is absent; a remake names those who take part, and a voter votes while it isn't gone. Any node of a voter takes a proposal up; every node judges a record against the votes in its history; a proposal is decided at a record that has two thirds of them in its history. A node takes up nothing lower than it took up, the highest number holds, and what hangs short of records shows and is ended by a later decision. A change is judged by what is decided in its history; the registry is resolved by what holds. * What is decided: numbers, the policy, first keys, the names of a user that is out, whose a contested name is, the voters' names, and a remake with the voters and the roots it names. * A remake: every node carries over what it holds beyond the base, whoever signed it, in its first changes and in order; the root change lists the digests of the base's statements, so nothing it left out is written again; decisions after the base are taken up again, and what a third of the voters' nodes held as gone, final or dropped stays so. An epoch's remakes are one matter. * A kind may name the characters of its names (lower case recommended); a kept name may be claimed, to stay through a change of policy. * The genesis change may hold users, without keys too, so a registry moved in has its users from the first change. * How often nodes stamp, a member entry's size, and what the genesis change sets for the swarm as a whole are numbers of the policy. * A node compares what it resolves with the stamps of the voters' nodes, and stops writing only where a third of them differ. Two wire versions spoken at once is a change for every swarm. * Stages: versions before remakes; the registry stage is for trying, the last for relying on it. 110 attacks.
* federation: all users are those there when a remake is proposedtv5 days1-36/+46
| | | | | | | | | | | | | | | | | | | | | | | A remake that names all users names those that count in its proposal's history, and still do at its base: the list is fixed with the proposal, so those who vote see it, and a user brought in while the votes run isn't on it. A fork's voters go the same way. And what reading the design from end to end found: * a remake does two things, not one: it throws history away and names who votes next; * nothing plain is written but a node's member entry and its records of proposals, said where namespaces are, too; * a node doesn't sync under a policy it doesn't know, whatever its version says; * two sides that only resolve differently show by their stamps; * on the wire a federated swarm's handshake says which epochs a node keeps, and a version speaks the one before its own: "no message changes" holds for a swarm that isn't federated; * three more things assumed: what was seen for long kept as the node goes, the cost of a digest with every stamp, two wire versions spoken at once.
* federation: voters named at a remake, versions, what is uncoveredtv5 days1-127/+216
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | What the maintainer asked of the design: * Who votes is named by the remake's proposal: all users at its base, as recommended, or a list. So those who vote say who votes next, and nobody votes by a rule of who created it. What that costs is said with it: every voter named counts toward what a decision takes, and all users is whoever brought users in. * What a policy no longer covers is uncovered at once where the decision holds: shown aside by every node, taken away by its user only, changed by nobody, and dropped once it has been seen there for `uncovered` (a number; 0: at once). Within an epoch a rule moves nothing in the document; the remake moves it. * Being seen takes nothing but a node that is up: its stamp is renewed about daily whatever else is written. A node left running keeps its voter present; naming the voters deals with that. * A release of a vote is best effort. * Continuum changing under a swarm: versions agree by what they compute, not by a number. Every stamp carries the digest of the snapshot its node would build there; a remake is taken up by the ID each node built; a node checks itself against the swarm before it writes with a new version; a change of the rules is a change of policy, and a node knows a policy by everything in it; recorded vectors every version passes. What can't change, what changes freely, and when it is a fork. * Who may connect is outside the design; the swarm key is a note. The attacks are 84. future-work.md: several roots, present by the user's own signature, a guide to moving to another swarm.
* federation: lines, time from the stamps, the last decision holdstv5 days1-493/+624
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | What the maintainer asked of the design, and what it changed: * Lines. A user is in while it has a line: a chain of creations and vouches from the root. When its creator goes, one that another vouched for stays without doing anything. The witnesses of its claims are the users on the lines it stands on: its creation's, and those of vouches it accepted (`accept`), so nobody gets a say about a user's names by vouching for it unasked. Why only its signer takes a vouch back is said where unvouching is. * Lengths of time are read between two stamps of one node, with what that node had seen at each: a voter is absent once a third of the voters saw it silent for long, a user gone once they saw it out. No clock is set against another, and why a third. * Decisions. A proposal is decided as soon as the nodes of two thirds of the voters hold its votes: at once, not at the end of a window. Any number are open at a time, within a limit a voter. Of the decisions of one matter the last holds, by which proposal was made in sight of which, so two decided at once stop nothing. A remake's base is named by its coordinator, a node of the proposer. * The policy changes by a decision, as a whole, with no remake: a remake throws history away and does nothing else. What a policy no longer covers doesn't count, and a remake keeps it aside for an epoch or drops it, as the policy says. Going apart by a fork stays for those a decision leaves behind. * What a remake leaves out, so that the document's size goes by what stands: a release is a statement in its epoch, the remake deletes. * Names are written as they are used, in the document and in the rules alike, and compared by their segments: structure, not meaning. A rule may leave a name to the voters to give. * Who vouched for whom changes nothing in who connects to whom: a node keeps its seeds and its user's other nodes as neighbours. The attacks are 83, the stages fifteen. future-work.md: guides to write (keys, the root, starting a swarm, voting, a registry), policies merged.
* federation: the fractions are numbers, a decision can be refusedtv5 days1-121/+169
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | What the maintainer answered to the design's last two questions, and what follows from it: * The fractions are numbers of the policy like the others, changed by a decision to any value: the share of the voters present a decision takes (two thirds recommended), the share of all of them (half), and the share whose stamps make the swarm's time and a user gone (a third). What is promised of decisions is argued for the recommended ones; Why it holds says what other values carry. * Refusing a decision: a node follows every decision that counts unless its operator has it refuse, by bounds on the policy's numbers or by hand. It then moves to the epoch built at the decision's history without it, which is the same for every node that refuses the same decision: those who think alike go on together without arranging it. While a proposal's votes run, a node whose bounds it breaks says so. A remake or a change of policy is refused by staying. * A decision gives a user name with its namespace, as designed. * No open question is left. The options not taken that the questions listed are among the roads not taken: first users voting as soon as created, voters let in without a remake, other reaches of an unvouch, a gone user coming back, first keys by those who vouch, a registry of only what is signed, numbers the design fixes. The attacks are 78; the stages name them by their numbers. future-work.md: bounds on more than numbers.
* federation: numbers without constraint, names by a decisiontv5 days1-171/+164
| | | | | | | | | | | | | | | | | | | | | | | | | | What the maintainer answered to the questions the design had left: * Every quantity of the design is a number of the policy, in one table with what the design recommends (the times, the limits, how much a remake's base may lack, how long the root's node lets a claim wait). The genesis policy sets each, and a decision changes any to any value: nothing constrains them. The fractions aren't numbers. * While a user is out, the voters free its names or give them to another by a decision, whatever `gone` says; a policy without `gone` has the names wait until a vouch or a decision. * Nobody is put out by a vote, said where unvouching is, with what there is against a user that harms within the rules. * A node keeps the epochs it left for as long as its operator says, `archiveRetention` at least, and says which; so a node coming back, or one joining from the genesis root, walks every epoch where an archive is within reach, and shows where the walk doesn't end at its peers' epoch. * Settled, and listed so: who votes, from where an unvouch undoes, gone for good, first keys by a decision, what the root holds unsigned. Two questions remain: whether the fractions are numbers too, and whether a decision gives a user name with its namespace. future-work.md: an archive of every epoch.
* federation: the answers to its open questionstv5 days1-218/+387
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | What the maintainer answered to the design's open questions, in the design: * The numbers. The limits and the times are the policy's numbers, set in the genesis policy and changed by a decision: 64 keys, 64 nodes and 256 addresses a user for a start; `window`, `absent`, `gone`, `archiveRetention`. A lowered limit takes nothing away. * A user nobody vouches for is out, and its names wait for it: nobody else claims them. Out for 30 days, as the stamps of a third of the voters show, it is gone: its names are free, and no vouch brings it back. So whoever relied on a vouch that an unvouch or a cut took back has that long to find another, and loses nothing meanwhile. * First keys by a decision. The voters are above every user's first keys: they give a user that lost them new ones, with a cut for the old. The promises say so, and what isn't promised: a name against two thirds of the voters. * What a decision is made of (a proposal, a vote, its release, the decision) is taken in any node's change, so a user without a node votes at whatever node it reaches, or by a vote carried to one. Such a user names another party's node to carry for it. * Coming back after remakes: by the archives on nobody's word, epoch by epoch, as far as peers keep them; else by an ID the operator gives, never because peers advertise an epoch. A statement is carried over from the epoch before only; older ones are quarantined for the user to sign again. * Resolving reads the decisions first: nothing below changes them. * The questions that remain are put with their options: who has a vote and from when, whether the voters put a user out against those who vouch for it, from where an unvouch undoes, whether gone is for good, the price of first keys by a decision, and what the root holds unsigned. The attacks are 76; the stages name them by their numbers. future-work.md: barring a user by a decision, and voting made easy for a user without a node.
* federation: a design for a swarm kept by parties that don't trust each othertv5 days2-0/+1255
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | federation.md designs what the krebs registry needs of continuum: a swarm several parties keep, any of them possibly malicious. Nothing of it is built. It is a mode a swarm is started in: one that isn't federated stays as it is. * The parties, what an adversary can do, and what is promised to an honest node: only valid changes; the same document and the same resolved registry for the same changes; values signed by their users; one holder per name; a claim nobody takes once the root's node saw it; no waiting for anyone to write; and, with fewer than a third of the voters present malicious, decisions by two thirds of them. And what isn't: one history under a partition, a name never contested, decisions against a third, against keys stolen later, against a malicious voter on both sides of a long partition, or with half the voters gone; a swarm no forked actor splits. * The design, on the strict policy. The policy is in the root. Users come by statements that only add; a user is its creation, its keys a tree, its nodes those it names. Vouching and the line; unvouching, with what exactly it takes back. A namespace per user. Names of declared kinds claimed, with rules and limits as data, some kept for a user without a claim. Of two claims out of each other's sight, one is late where its own side can be shown to have seen the other first; where that leaves not exactly one, the name is contested, and a release gives it up. Cuts: what a key signed outside a set of changes stops counting, whenever it shows up; a key is cut off by one above it only. Bans change what those who follow read, not what anyone takes. * Decisions: votes signed by the voters, over days, by numbers the document holds and a decision changes. Who votes is in an epoch's root change: the root alone at first, then, with each remake, the users a root created, so nobody fills the votes without two thirds of those there agreeing. A time read from the document, by which those away for long don't count; a decision taken up by the voters' nodes while its votes run, and counting with two thirds of those present and more than half of all, so that honest voters never make two however long a partition lasts. The root's seat given anew by a decision, and what a root finds when it comes back. Remakes: the base is the decision's history, every node builds the snapshot from the registry as it resolves, statements are carried over, and what counted at the base is final. A change of policy as a fork each operator follows or not, as Bitcoin changes its rules; it may name who votes on its side, and is the way out for those a decision leaves behind. * What every node judges and how the registry resolves, how a partition shows and how a side comes back, links along the line, tokens of one's own party only. Why the promises hold, the roads not taken, what members are trusted with then, what is continuum's and what the application's, 69 attacks with the tests that show them, the registry's layout with moving in and writing out, fourteen stages, and ten open questions with a choice for each. future-work.md gets what the design leaves.
* screen: `Screening::of`'s doc on `of`tv27 hours1-1/+1
| | | | | `of`'s doc line sat above `policed`'s, as its first line, and `of` had none.
* screen: a test's actor named through `document`tv28 hours1-1/+1
| | | | | | The test of a value hiding the policy's place made its actor through `automerge::` outside `src/document`, which the test that Automerge stays behind `document` refuses. It names `document::ActorId`.
* screen: what is at the policy's place found in every map on the waytv28 hours2-5/+87
| | | | | | | | | | | | | | | | A node judges what it receives once anything is at the policy's place (`Screening::policed`), as a read shows it. But a read shows the newest of the values written at once at a key, merging maps only where that one is a map: a string written at `_continuum/access` beside the map holding the policy, and newer, hides it. So the policy's place read as empty, a change putting `"signed"` into the hidden policy went unjudged where it came alone, and deleting the string left a strict policy no change was refused for. Whether anything is at the policy's place is asked of every map of each key on the way to it, read or not (`store::held_anywhere`), as `store::by_seq` walks them. A swarm with nothing there is screened as before.
* docs: the strict remake among the kindstv35 hours2-3/+4
| | | | | | The glossary and the `kind` row of `remakes.md` named three kinds of remake; `strict` is the fourth, and takes `baseHeads` as `replace` and `import` do.