Pragmatic signature aggregation with BLS

Sharding
4

Good question :slight_smile: It’s the same beacon chain as for sharding. We are considering merging Casper and sharding such that the Casper FFG system logic and the sharding system logic both run on the beacon chain instead of on the legacy main chain.

For validators to make 32 ETH deposits the legacy chain would have a one-way burn contract. Withdrawals of deposits and rewards would remain disabled until we ship the EVM 2.0 in the shards. Advantages of this new approach include:

  • Segregation: There would be a clean segregation between system logic and application logic. This makes it easier to upgrade the system logic without upsetting dapps. It also means that system contracts won’t have to compete with dapps on gas.
  • Future-proofing: Eventually the EVM as we know it (EVM 1.0) will be deprecated and replaced by EVM 2.0 in the shards, at which point the beacon chain will become the root chain. As such, we want to avoid writing long-term system logic in EVM 1.0. Also full PoS would likely get shipped earlier.
  • Unity: There would be a single version of proof-of-stake instead of one for Casper and one for sharding. This allows for capital reuse (all validators both Casper and sharding validators), a unified Ethereum 2.0 design, and closer cooperation between the Casper and sharding teams.
  • Simplicity: Classes of issues go away, such as dealing with gas and writing code in Solidity. While withdrawals are disabled the manager contract on the beacon chain will be “unhackable”, and much easier to write and deploy.

Good point—I’ll look into it. Do you have a link to the patent for the original BLS scheme?

We can use an aggregate signature, it just means that a fresh public key needs to computed for every new subset of signing validators (as opposed to always reusing the same public key). That’s OK because computing public keys only involves fast multiplications and is highly parallelisable.

The new BLS paper does also provide a threshold signature scheme but it requires a one-time interactive setup to build the public key. This setup doesn’t scale well and doesn’t easily allow for changes to the validator set.

8 Likes