[Vitalik: The scalability layers in blockchain can be divided into computation, data, and state] Foresight News reports that Vitalik Buterin tweeted, 'The scalability layers in blockchain can be divided into computation, data, and state. Computation is easier to scale than data. You can parallelize computation, require code block builders to provide various "hints," or directly replace any amount of computational data with proofs. Data is the key. If data availability guarantees are needed, they must be provided—there is no alternative. However, the PeerDAS solution can be used to partition and encode it into erasure codes. It can also achieve graceful degradation: if a node's data capacity is only 1/10 of other nodes, the block size it generates will always be 1/10. State management is the most difficult. To ensure the ability to verify a single transaction, complete state is required. If a tree structure is used to replace the state while retaining the root node, complete state is still needed to update the root node. Although there are partitioning solutions, they involve architectural changes and are inherently not universal. Therefore, if state can be replaced with data (without introducing new forms of centralization), this approach should be seriously considered by default. Similarly, if computation can be replaced with data (without introducing new forms of centralization), this approach should also be seriously considered by default.'