Lose one provider. Keep the backup.

NostrBorg keeps Borg repositories as encrypted, content-addressed blobs across independent replicas. The product test is simple: destroy one store, restore byte-identical ciphertext from the survivor.

Opaque by default

Storage providers see SHA-256 blobs and sizes, not file names, snapshots, or plaintext. Recovery keys stay on the operator’s device.

Replica, then shard

Full-replica loss is the first proof. Sharding and erasure coding wait until full replication works and users reject the storage cost.

Local Borg still wins

This hosted drill proves the replica/CAS contract. The repository still uses Borg for chunking, encryption, and `borg check --verify-data`.

Production restore drill

Destroy one replica. Restore from the other.

This runs synthetic ciphertext only. It uploads opaque blobs to two in-process replicas, deletes replica A, reassembles from replica B, and byte-compares the result. No plaintext and no real Borg repository leave this request.

No drill has been run in this session yet.