Skip to content
Switch branches/tags

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time

Summa Relay

This is a Bitcoin Relay. It uses 1 + 1/n slots per header relayed (n is currently 4), and 2 slots to externalize useful information (best chain tip and best shared ancestor of latest reorg).

Implementations are available in Solidity (for EVM chains) and Golang using the cosmos-sdk framework.

How does it work?

The core idea behind the relay is to minimize storage costs by increasing calldata costs. Rather than storing headers, the relay stores the hashPrevBlockfield of each header and the height of every nth header. Should the relay need to reference information in old headers (like the difficulty), the header data is passed to the relay again, and validated against known hashPrevBlock links. This allows the relay to check that newly submitted blocks are valid extensions of existing blocks, without storing all past header information.

As opposed to other relays, we separate the function of the relay into two categories: "learning about new blocks" and "following the best chain tip." Users may add new blocks in groups of at least 5 by calling addHeaders If the block slice includes a difficulty retarget, users are required to call addHeadersWithRetarget, which performs additional validation. The relay does not update its tip unless it is specifically requested to do so by a user. The user must call markNewHeaviest with the new heaviest, the old heaviest header, and the digest of their most recent common ancestor (which may be the old heaviest header.

As part of the process, the relay externalizes the most recent common ancestor, which is to say, the heaviest header that both old and new heaviest tip confirm. This is a metric of "subjective finality" for that block. During normal operation without reorgs it lags behind the tip by 5 blocks. During reorgs, it is the shared base of the competing branches (and as such may move backwards!). This indicates that competing sets of miners both viewed it as subjectively finalized. As such, it is a reasonable source of finalization information for relay-consuming smart contracts.

This model provides large gas savings compared to previous relay designs (TODO: benchmarking). It also gets especially attractive if EIP2028 activates, reducing calldata gas costs.

A Note on Endianness

Bitcoin internally uses little-endian representations of integers and digests. Block explorers and most user-facing applications use the more common big-endian representation. To minimize order swaps and prevent confusion, all our tooling uses the LE representation exclusively. If using the JS, rust, golang, or python tooling in bitcoin-spv, everything will Just Work. If writing custom software using data from block explorers, full nodes, or other data sources, make sure digests are LE before submitting to the relay.

Requests and Proofs

The Relay implementations here have an SPV request system built in. This allows for abstraction of the off-chain proving software. Requesters don't need to write a custom Bitcoin indexer, and existing Bitcoin indexers can work with any requester, whether it's a module, a smart contract, or a user.

The relay coordinates an interaction between 3 roles:

  1. Requester: creates a new SPV Proof request and designates a Handler
  2. Handler: handles incoming SPV Proofs on the Requester's behalf
  3. Indexer: watches requests, indexes Bitcoin, and provides SPV Proofs

While implementation details differ, the architecture is simple:

  1. Requesters register a request for SPV Proofs.
    1. The request specifies a transaction filter and a proof handler.
    2. golang: submit a MsgRequestProof.
    3. golang CLI: relaycli tx relay newrequest.
    4. solidity: OndemandSPV.request().
  2. An event with request details is logged.
    1. golang: watch for proof_request events.
    2. solidity: subscribe to NewProofRequest events.
  3. Indexers watch the Bitcoin chain for transactions that satisfy Requests.
    1. Example.
  4. Indexers create an SPV Proof and submit it to the relay.
    1. golang: submit a MsgProvideProof.
    2. golang CLI: relaycli tx relay provideproof.
    3. solidity: call OnDemandSPV.provideProof().
  5. The relay validates this proof.
  6. If valid, on-chain handler dispatches tx info to the proof Handler
    1. golang: the module's ProofHandler routes info the the Handler
    2. solidity: the relay calls spv() on the handling contract

Essentially the requester is subscribing to a feed of Bitcoin transactions matching a specific filter. This filter can specify which UTXO is being spent, and/or an address that receives funds. The handler expects to receive a stream of transactions that meet the filter's specifications.

Note: Due to solidity constraints, this filter system is unrelated to existing Bitcoin filtering systems (e.g. BIP37 & BIP157) In the future, the filter system may be upgraded to support more complex transaction descriptions.

Important: All requests may be filled more than once. Setting a spends filter is NOT sufficient to prevent this, as long reorgs may cause a UTXO to be spent multiple times. There is NO WAY to ensure that only a single proof is provided, so the handler should deal with multiple proofs gracefully.

Misc Project Notes

Complete relays are available in Solidity, for EVM-based chains (like Ethereum) and Golang using the cosmos-sdk framework.

The Python relay mainter in ./maintainer/ is not thoroughly tested, and does not yet support the cosmos-sdk relay.