Q: Why can't rollups just fork all of the parallelization related client level changes we've made for Sei v2?
A: Bandwidth constraints
All of the code we're writing for Sei v2 is open sourced, so ethereum clients can theoretically fork sei and try using optimistic parallelization.
Unfortunately, parallelization on eth rollups doesn’t work due to bandwidth constraints. Each byte of data that is written to eth l1 currently costs 16 gas, and compressed eth transfers need 12 bytes of data (
x.com/jayendra_jog/status/15……).
This results in a theoretical upper bound of ~6k TPS assuming all of the eth L1 block space is used for writing compressed tx data from rollups. In practice, you see much lower than that bc there ends up being a lot of contention for L1 block space (uniswap, opensea, blur all contend for block space with arbitrum, optimism).
This isn’t something that protodank sharding will solve either, since the amount of data that can be written onchain remains roughly the same (
x.com/jayendra_jog/status/16……).
If eth L2s try to get greater throughput via optimistic parallelization, they wouldn’t be able to write all of the tx data to the base layer.