Are you interested in RIP-7728 support? I find this one perspective.
https://github.com/0xurb/l1sload - example implementation on revm (release/v45).
When optimism flow will be finished, we can do more exciting precompiles, that use state and L1 context.
Risks and open questions:
Can we require L2 nodes to rely on L1 archive nodes?
[ ] TBD
How to deal with RPC errors?
[ ] TBD
Impact of RPC latency on sequencing? How do we price this in gas?
[x] some latency (tentative) cost for used gas calculation is included per slot
Can we guarantee that the accessed L1 state is recent?
[ ] TBD
Should we allow state reads at arbitrary L1 block height?
[x] state has an access through L1Block.sol. And if EVM context storage is returns error, we can take BlockId::Latest on RPC calls.
Are you interested in RIP-7728 support? I find this one perspective. https://github.com/0xurb/l1sload - example implementation on revm (release/v45). When optimism flow will be finished, we can do more exciting precompiles, that use state and L1 context.
Risks and open questions:
Can we require L2 nodes to rely on L1 archive nodes?
How to deal with RPC errors?
Impact of RPC latency on sequencing? How do we price this in gas?
Can we guarantee that the accessed L1 state is recent?
Should we allow state reads at arbitrary L1 block height?
BlockId::Lateston RPC calls.