ssb-db

secret-stack plugin which provides storing of valid secure-scuttlebutt messages in an append-only log.

What does it do?

ssb-db provides tools for dealing with unforgeable append-only message feeds. You can create a feed, post messages to that feed, verify a feed created by someone else, stream messages to and from feeds, and more (see API).

"Unforgeable" means that only the owner of a feed can modify that feed, as enforced by digital signing (see Security properties). This property makes ssb-db useful for peer-to-peer applications. ssb-db also makes it easy to encrypt messages.

Example

In this example, we create a feed, post a signed message to it, then create a stream that reads from the feed.

/**
 * create an ssb-db instance and add a message to it.
 */

//create a secret-stack instance and add ssb-db, for persistence.
var createApp = require('secret-stack')({})
  .use(require('ssb-db'))


// create the db instance.
// Only one instance may be created at a time due to os locks on port and database files.

var app = createApp(require('ssb-config'))

//your public key, the default key of this instance.

app.id

//or, called remotely

app.whoami(function (err, data) {
  console.log(data.id) //your id
})

// publish a message to default identity
//  - feed.add appends a message to your key's chain.
//  - the `type` attribute is required.

app.publish({ type: 'post', text: 'My First Post!' }, function (err, msg, hash) {
  // the message as it appears in the database:
  console.log(msg)

  // and its hash:
  console.log(hash)
})

// stream all messages for all keypairs.
pull(
  app.createLogStream(),
  pull.collect(function (err, ary) {
    console.log(ary)
  })
)

// stream all messages for a particular keypair.
pull(
  app.createHistoryStream({id: app.id}),
  pull.collect(function (err, ary) {
    console.log(ary)
  })
)

Concepts

Building upon ssb-db requires understanding a few concepts that it uses to ensure the unforgeability of message feeds.

Identities

An identity is simply a public/private key pair.

Even though there is no worldwide store of identities, it's infeasible for anyone to forge your identity. Identities are binary strings, so not particularly human-readable.

Feeds

A feed is an append-only sequence of messages. Each feed is associated 1:1 with an identity. The feed is identified by its public key. This works because public keys are unique.

Since feeds are append-only, replication is simple: request all messages in the feed that are newer than the latest message you know about.

Note that append-only really means append-only: you cannot delete an existing message. If you want to enable entities to be deleted or modified in your data model, that can be implemented in a layer on top of ssb-db using delta encoding.

Messages

Each message contains:

A message object. This is the thing that the end user cares about. If there is no encryption, this is a {} object. If there is encryption, this is an encrypted string.
A content-hash of the previous message. This prevents somebody with the private key from changing the feed history after publishing, as a newly-created message wouldn't match the "prev-hash" of later messages which were already replicated.
The signing public key.
A signature. This prevents malicious parties from writing fake messages to a stream.
A sequence number. This prevents a malicious party from making a copy of the feed that omits or reorders messages.

Since each message contains a reference to the previous message, a feed must be replicated in order, starting with the first message. This is the only way that the feed can be verified. A feed can be viewed in any order after it's been replicated.

Object ids

The text inside a message can refer to three types of ssb-db entities: messages, feeds, and blobs (i.e. attachments). Messages and blobs are referred to by their hashes, but a feed is referred to by its signing public key. Thus, a message within a feed can refer to another feed, or to a particular point within a feed.

Object ids begin with a sigil @ % and & for a feedId, msgId and blobId respectively.

Note that ssb-db does not include facilities for retrieving a blob given the hash.

Replication

It is possible to easily replicate data between two instances of ssb-db. First, they exchange maps of their newest data. Then, each one downloads all data newer than its newest data.

ssb-server is a tool that makes it easy to replicate multiple instances of ssb-db using a decentralized network.

Security properties

ssb-db maintains useful security properties even when it is connected to a malicious ssb-db database. This makes it ideal as a store for peer-to-peer applications.

Imagine that we want to read from a feed for which we know the identity, but we're connected to a malicious ssb-db instance. As long as the malicious database does not have the private key:

The malicious database cannot create a new feed with the same identifier
The malicious database cannot write new fake messages to the feed
The malicious database cannot reorder the messages in the feed
The malicious database cannot send us a new copy of the feed that omits messages from the middle
The malicious database can refuse to send us the feed, or only send us the first N messages in the feed
Messages may optionally be encrypted. See test/end-to-end.js.

API

SecretStack.use(require('ssb-db')) => SecretStackApp

Adds ssb-db persistence to a secret-stack setup. Without other plugins, this instance will not have replication or querying. Loading ssb-db directly is useful for testing, but it's recommended to instead start from a plugin bundle like ssb-server

Because of legacy reasons, all the ssb-db methods are mounted on the top level object, so it's app.get instead of app.db.get as it would be with all the other ssb-* plugins. In the API docs below, we'll just call it db

db.get (id | seq | opts, cb)

Get an ssb message. If id is a message id, the message is returned. If seq is provided, the message at that offset in the underlying flumelog is returned. If opts is passed, the message id is taken from either opts.id or opts.key. If opts.private = true the message will be decrypted if possible. If opts.meta = true is set, or seq is used, the message will be in {key, value: msg, timestamp} format. Otherwise the raw message (without key and timestamp) are returned. This is for backwards compatibility reasons. Given that most other apis (such as createLogStream) by default return {key, value, timestamp} it's recommended to use get({id: key, meta: true}, cb)

db.add(msg, cb)

append a raw message to the local log. msg must be a valid, signed message. ssb-validate is used internally to validate messages.

db.publish(content, cb)

create a valid message with content with the default identity and append it to the local log. ssb-validate is used to construct a valid message.

db.whoami(cb)

call back with the default identity for the db.

db.createLogStream({lt,lte,gt,gte: timestamp, reverse,old,live,raw: boolean, limit: number}) => PullSource

create a stream of the messages that have been written to this instance in the order they arrived. This is mainly intended for building views. The objects in this stream will be of the form:

{
  key: Hash, value: Message, timestamp: timestamp
}

timestamp is generated by monotonic-timestamp when saving the message.

gt, gte, lt, lte ranges are supported, via ltgt if reverse is set to true, results will be from oldest to newest. if limit is provided, the stream will stop after that many items. old and live return wether to include old and live (newly written messages) as via pull-live

if raw option is provided, then instead createRawLogStream is called.

db.createRawLogStream(lt,lte,gt,gte: offset, reverse,old,live: boolean, limit: number})

provides access to the raw flumedb log. ranges refer to offsets in the log file.

messages are returned in the form:

{
  seq: offset,
  value: {key: Hash, value: Message, timestamp: timestamp}
}

all options supported by flumelog-offset are supported.

db.createHistoryStream({id: feedId, seq: int?, live: bool?, limit: int?, keys: bool?, values: bool?}) -> PullSource

Create a stream of the history of id. If seq > 0, then only stream messages with sequence numbers greater than seq. if live is true, the stream will be a live mode

Note: since createHistoryStream is provided over the network to anonymous peers, not all options are supported. createHistoryStream does not decrypt private messages.

db.messagesByType({type: string, live,old,reverse: bool?, gt,gte,lt,lte: timestamp, limit: number }) -> PullSource

retrieve messages with a given type. All messages must have a type, so this is a good way to select messages that an application might use. Returns a source pull-stream.

as with createLogStream messagesByType takes all the options from pull-level#read (gt, lt, gte, lte, limit, reverse, live, old)

ranges may be a timestamp, of the local received time.

db.createFeedStream(lt,lte,gt,gte: timestamp, reverse,old,live,raw: boolean, limit: number})

like createLogStream, but messages are in order of the claimed time, instead of the received time. This may sound like a much better idea, but has surprising effects with live messages - you may receive a old message in real time - but for old messages, it makes sense.

all standard options are supported.

db.createUserStream ({id: feed_id, lt,lte,gt,gte: sequence, reverse,old,live,raw: boolean, limit: number})

createUserStream is like createHistoryStream, except all options are supported. Local access is allowed, but not remote anonymous access. createUserStream does decrypt private messages.

db.links({source: feedId?, dest: feedId|msgId|blobId?, rel: string?, meta: true?, keys: true?, values: false?, live:false?, reverse: false?}) -> PullSource

Get a stream of links from a feed to a blob/msg/feed id.

The objects in this stream will be of the form:

{ source: feedId, rel: String, dest: Id, key: MsgId, value: Object? }

source (string, optional): feed id..
dest (string, optional): An id or filter, specifying where the link should point to. To filter, just use the sigil of the type you want: @ for feeds, % for messages, and & for blobs.
rel (string, optional): Filters the links by the relation string.

If opts.values is set (default: false) value will be the message the link occurs in. If opts.keys is set (default: true) key will be the message id. If opts.meta is unset (default: true) source, hash, rel will be left off.

Note: if source, and dest is provided, but not rel, ssb will have to scan all the links from source, and then filter by dest. your query will be more efficient if you also provide rel.

db.addMap(fn)

Add a map function to be applied to all messages on read. The fn function is should expect (msg, cb), and must eventually call cb(err, msg) to finish.

These modifications only change the value being read, but the underlying data is never modified. If multiple map functions are added, they are called serially and the msg output by one map function is passed as the input msg to the next.

Additional properties may only be added to msg.value.meta, and modifications may only be made after the original value is saved in msg.value.meta.original.

db.addMap(function (msg, cb) {
  if (!msg.value.meta) {
    msg.value.meta = {}
  }

  if (msg.value.timestamp % 3 === 0)
    msg.value.meta.fizz = true
  if (msg.timestamp % 5 === 0)
    msg.value.meta.buzz = true
  cb(null, msg)
})

const metaBackup = require('ssb-db/util').metaBackup

db.addMap(function (msg, cb) {
  // This could instead go in the first map function, but it's added as a second
  // function for demonstration purposes to show that `msg` is passed serially.
  if (msg.value.meta.fizz && msg.value.meta.buzz) {
    msg.meta = metaBackup(msg.value, 'content')

    msg.value.content = {
      type: 'post',
      text: 'fizzBuzz!'
    }
  }
  cb(null, msg)
})

db._flumeUse(name, flumeview) => view

Add a flumeview to the current instance. This method was intended to be a temporary solution, but is now used by many plugins, which is why it starts with _.

see undocumented creating a secret-stack plugin.

db.getAtSequence([id, seq], cb(err, msg))

get a message for a given feed id with given sequence. calls back a message or an error, takes a two element array with a feed id as the first element, and sequence as second element.

needed for ssb-ebt replication

db.getVectorClock(cb)

load a map of id to latest sequence ({<id>: <seq>,...}) for every feed in the database.

needed for ssb-ebt replication

db.progress

return the current status of various parts of the scuttlebut system that indicate progress. This api is hooked by a number of plugins, but ssb-db adds an indexes section. (which represents how fully built the indexes are)

the output might look like:

{
  "indexes": {
    "start": 607551054,
    "current": 607551054,
    "target": 607551054
  },
}

progress is represented linearly from start to target. Once current is equal to target the progress is complete. start shows how far it's come. The numbers could be anything, but start <= current <= target if all three numbers are equal that should be considered 100%

db.status

returns metadata about the status of various ssb plugins. ssb-db adds an sync section, that shows where each index is up to. output might took like this:

{
  "sync": {
    "since": 607560288,
    "plugins": {
      "last": 607560288,
      "keys": 607560288,
      "clock": 607560288,
      "time": 607560288,
      "feed": 607560288,
      "contacts2": 607560288,
      "query": 607560288,
      ...
    },
    "sync": true
  }
}

sync.since is where the main log is up to, since.plugins.<name> is where each plugin's indexes are up to.

db.version

return the version of ssb-db. currently, this returns only the ssb-db version and not the ssb-server version, or the version of any other plugins. We should fix this soon

db.queue(msg, cb)

add a message to be validated and written, but don't worry about actually writing it. the callback is called when the database is ready for more writes to be queued. usually that means it's called back immediately.

not exposed over RPC.

db.flush(cb)

callback when all queued writes are actually definitely written to the disk.

db.post(fn({key, value: msg, timestamp})) => Obv

observable that calls fn whenever a message is appended (with that message)

not exposed over RPC.

db.since(fn(seq)) => Obv

an observable of the current log sequence. This is always a positive integer that usually increases, except in the exceptional circumstance that the log is deleted or corrupted.

db.addUnboxer({key:unboxKey, value: unboxValue})

add an unboxer object, any encrypted message is passed to the unboxer object to test if it can be unboxed (decrypted)

db.unbox(data, key)

attempt to decrypt data using key. Key is a symmetric key, that is passed to the unboxer objects.

deprecated apis

db.getLatest(feed, cb(err, {key, value: msg}))

get the latest message for the given feed, with {key, value: msg} style.

maybe used by some front ends, and by ssb-feed.

db.latestSequence(feed, cb(err, sequence))

callback the sequence number of the latest message for the given feed. (I don't think this is used anymore)

db.latest() => PullSource

returns a stream of {author, sequence, ts} tuples. ts is the time claimed by the author, not the received time. (I don't think this is used anymore)

db.createWriteStream() => PullSink

create a pull-stream sink that expects a stream of messages and calls db.add on each item, appending every valid message to the log.

db.createFeed(keys?) => Feed (deprecated)

deprecated: use ssb-identities instead

may only be called locally, not from a ssb-client connection.

Create a Feed object. A feed is a chain of messages signed by a single key (the identity of the feed). This handles the state needed to append valid messages to a feed. If keys are not provided, then a new key pair will be generated.

The following methods apply to the Feed type.

Feed#add (message, cb) (deprecated)

Adds a message of a given type to a feed. This is the recommended way to append messages. message is a javascript object. It must be a {} object with a type property that is a string between 3 and 32 chars long.

If message has recps property which is an array of feed ids, then the message content will be encrypted using private-box to those recipients. Any invalid recipients will cause an error, instead of accidentially posting a message publically or without a recipient.

Feed#id (deprecated)

the id of the feed (which is the feed's public key)

Feed#keys

the key pair for this feed.

Stability

Stable: Expect patches, possible features additions.

License

MIT

gmarcos87/ssb-db