What data does the Polkadot relay chain really need to store for parachain block validation?

Question

I understand that the relay chain needs to store proofs for a certain amount of final blocks if a dispute occurs. But it's not clear to me what the implementation decisions were for explicitly writing certain things to Polkadot's runtime storage (such as the bit fields for availability and the onChainVotes from the para_inherent pallet).

Why is writing this to Polkadot's runtime storage really necessary if it's only needed per block? Assuming there's a good reason, why can't it work outside the runtime?

Answer 1

For bitfields: I am pretty sure that the only reason we store them is so we can accumulate them over several blocks until we reached availability/cores time out. While validators will re-send bitfields every block, this adds some more resilience in case of spurious network problems for example.

For the onChainVotes stuff: This is an implementation detail. It is strictly speaking not needed to store this on chain, but it is useful for disputes to have that information easily accessible by the node. Easily accessible means a versioned runtime API, hence the need to put it in storage.

For disputes it is crucial to have backing votes available for a candidate in case a dispute is raised at some point. We use the onChainVotes stuff in the dispute-coordinator to scrape votes from chain state at every block, as that is less troublesome than reading block data directly on the node side. Because we read the state at each block, there is no need to preserve votes in state for longer.

Hope that makes sense.

Answer 2

You can find additional resources on this topic here:

• https://spec.polkadot.network/#chapter-anv
• https://w3f.github.io/parachain-implementers-guide

Answer 3

This article will answer the question in great details.

• https://polkadot.network/blog/polkadot-v1-0-sharding-and-economic-security/