Replication is a consensus rule
Three copies, enforced by the chain and reassigned deterministically. Not a service level agreement, a state transition.
03Storage subprotocol
SNIP
The ledger holds the proof. The mesh holds the bytes.
Summary
Storage Node Interface Protocol. The chain keeps merkle roots, access lists, and fee pools. File bytes live in a libp2p mesh of archives that earn by proving, every block, that they still hold the chunks they were deterministically assigned.
Role in the stack
Holds everything too large for state: credential payloads, course material, mail attachments, model weights.
Archives are challenged every block
Proof of Retrievability runs continuously. An archive that cannot answer in time is slashed.
No IPFS, no second token
Storage economics settle in Koppa directly at the consensus layer. There are no smart contracts in the storage path.
Architecture
- Chunk Client side, content addressed
- Commit Merkle root and access list on chain
- Assign Deterministic, 3 archives
- Challenge Proof of Retrievability, every block
- Pay or slash Fee pool settles in Koppa
Specification
Values are taken from the protocol repository and from live chain parameters.
SNIP specification
| Transport | libp2p with Circuit Relay v2 and DCUtR hole punching |
|---|---|
| Replication factor | 3, enforced at consensus |
| Integrity | Per-block Proof of Retrievability challenges |
| Failure handling | Stake slashing on missed challenge |
| Encryption keys | X25519, derived from the Ed25519 account key |
| Object lifecycle | Pending, Active, Abandoned |
| Storage fee | 100 base units per byte |
| Max chunks per file | 1,048,576 |
| Activation height | 5,200,000 |
| Archive platform | Linux x86_64 supported, macOS Apple Silicon experimental |
| Client platform | Linux, macOS, Windows via WSL2, ChromeOS via Crostini |
| Contributor app | Native macOS |
| License | MIT or Apache-2.0 |
Endpoints and source
- snip.sumchain.io app
- Source repo