PouchDB is a cross-platform database. It's possible to build a database, dump it, and then load it into another PouchDB instance, which is how we'd like to support client-side snapshots (it also supports real-time sync, which is nice for reactive frontends!)
The logic for resuming from a snapshot is:
Drop local databases (reset)
Download the database dump
Load this database dump.
Find the blockHash of the most recent transaction recorded
Restart all reducers from that latestBlockHash
or, when syncing
db.replicate.from(remote)
In this way, a client can download snapshots from trusted third parties (eventually we'll use a decentralized computation network to produce these indexes) and then resume following the index in real-time, locally, block-by-block talking directly to an Ethereum node.
The first step to supporting PouchDB snapshots is to add support to Gnarly's IPersistInterface for talking to PouchDB. By default PouchDB will use an indexeddb adaptor in browser and a leveldb adaptor on node, which is probably fine. If there are any inconsistencies, though, we can use the websql adaptor in browser and the pouchdb-adapter-node-websql adaptor in node (which ends up writing to disk via sqlite).
PouchDB is a cross-platform database. It's possible to build a database, dump it, and then load it into another PouchDB instance, which is how we'd like to support client-side snapshots (it also supports real-time sync, which is nice for reactive frontends!)
The logic for resuming from a snapshot is:
blockHashof the most recent transaction recordedlatestBlockHashor, when syncing
db.replicate.from(remote)In this way, a client can download snapshots from trusted third parties (eventually we'll use a decentralized computation network to produce these indexes) and then resume following the index in real-time, locally, block-by-block talking directly to an Ethereum node.
The first step to supporting PouchDB snapshots is to add support to Gnarly's
IPersistInterfacefor talking to PouchDB. By default PouchDB will use an indexeddb adaptor in browser and a leveldb adaptor on node, which is probably fine. If there are any inconsistencies, though, we can use thewebsqladaptor in browser and thepouchdb-adapter-node-websqladaptor in node (which ends up writing to disk via sqlite).Use pouchdb-server to persist data serverside.