Skip to main content
As a stream grows, replaying it from offset -1 becomes expensive. Snapshots solve this by storing a compacted representation of all data up to a certain offset. 
PUT /ds/b/s/snapshot/{offset}    — publish a snapshot
GET /ds/b/s/snapshot             — read the latest snapshot
Once a snapshot is published at a given offset, the server may discard all data before that offset. Reads to discarded offsets return 410 Gone, directing clients to use /bootstrap instead. Who creates snapshots? Typically the application. For instance, a CRDT-based editor periodically computes the merged document state and publishes it as a snapshot. This bounds the amount of data new clients need to process.