SSB partial replication

A collection of functions useful for replicating a part of the log
instead of everything. This is superseeded by SSB secure partial replication.

api

$3

The method has exactly the same interface as createHistoryStream
because it just wraps the function. It is much faster though. See
appendix for more information.

$3

This function does the same as getFeed except if one does
not specify seq the latest sequence for the feed with be fetched and
used. This allows one to get the latest X messages from a feed using
the limit option without knowing how many messages the feed
currently has.

$3

Get all the messages of a tangle
given a root message id. This can be used to fetch threads or similar
tangles. This is similar in spirit to
ssb-ooo except there is no protocol
involved to fetch messages from other nodes.

$3

Get messages of a particular type for a feed. This can be used to
fetch say all contact messages for a particular feed to construct the
social graph without having to download all the messages of the feed.

Appendix

There is something wrong with createHistoryStream over a network
connection. Locally it takes around 600ms, using net protocol the same
call takes 6.500ms and using ws the exact same call takes 20.000ms?

``javascript
var pull = require('pull-stream')

var remote = 'ws:between-two-worlds.dk:8989~shs:lbocEWqF2Fg6WMYLgmfYvqJlMfL7hiqVAV6ANjHWNw8=.ed25519'
//remote = 'net:between-two-worlds.dk:8008~shs:lbocEWqF2Fg6WMYLgmfYvqJlMfL7hiqVAV6ANjHWNw8=.ed25519'

require('ssb-client')({ remote }, (err, sbot) => {
if (err) throw err
console.time("downloading messages")
pull(
sbot.createHistoryStream({id: '@ye+QM09iPcDJD6YvQYjoQc7sLF/IFhmNbEqgdzQo3lQ=.ed25519', seq: 27000, keys: false}),
pull.drain((msg) => {
console.log(msg)
}, (err) => {
if (err) throw err

console.timeEnd("downloading messages")
sbot.close()
})
)
})
`

The problem seems to be buried in the
legacy
protocol for ssb-replication somewhere.

This module simply exposes
createHistoryStream
as partialReplication` without any of the legacy overhead and we are
back to 600ms again.