Layer two scaling solution store data out of storage, but they often have a need to bring it in a transaction for security purposes (fraud proofs, data availability, etc.). When data is brought on-chain, this also often includes large merkle proofs relating it back to concise on-chain state.
Currently the cost of updating storage is ~ 5000 gas. The cost of verifying a single Merkle tree of depth 32 is ~70 000 gas. It is cheaper to skip L2 scaling and just store on chain.
Examples of applications that are currently constrained by transaction data cost:
Final implementation is a trivial update of costs, but before that we need to extensively research what would be an appropriate gas cost and it’s impact on the network.
- Remco’s talk in Stanford https://www.youtube.com/watch?v=dHZiWFWCGNM&feature=youtu.be&t=68
- Vitalik’s talk in Tapei https://youtu.be/mOm47gBMfg8?t=755