On the Plasma chain, the basic structure of blocks is similar to that of a regular chain, with transactions and state changes completed within the child chain. However, from the original design intention, Plasma does not aim for the mainnet to handle these details.

After each new block is generated, the operator's task is simple—hash all transactions and state data within the block to generate a Merkle Root (some implementations use a transaction Root, others use a state or UTXO Root).

This Root is essentially:

"An unalterable fingerprint of the entire history up to this block."

This is the first layer of compression. No matter how many transactions are in the block, the mainnet only needs to see a fixed-length hash value.

Next comes the second layer of compression. Plasma doesn't rush to submit each block individually to the chain; instead, it accumulates the header information of multiple blocks and packages them into a higher-level Root. Ultimately, what the mainnet receives is usually not a single block header but a consolidated commitment to the Plasma state over a period of time.

What the mainnet sees at the end is very streamlined:
• Block number or time interval
• Corresponding Root value
• Necessary timestamp

The mainnet does not verify transactions or store data; it only retains these Roots as an immutable proof of time.

When someone wants to exit or raise a dispute, they need to provide transaction data and Merkle proofs to reconcile against these block headers.

No accounting during normal operation; only when issues arise do you do the accounting—this is the core logic behind why Plasma compresses block headers before submitting them to the mainnet.