Layer one and layer two: two ends of one scaling spectrum

A rollup executes transactions outside the base chain and posts data back to it. The Ethereum documentation describes two families, and the difference between them is the mechanism that decides whether the posted result is correct. An optimistic rollup assumes the result is correct and provides a window during which anyone can submit a fraud proof; if no valid challenge arrives, the result stands. A zero-knowledge rollup posts a validity proof alongside the result, and the base chain's contract verifies the proof before accepting the state update.

The two families therefore differ in what they assume and in when the assumption is resolved. An optimistic rollup's security depends on at least one honest party watching and being willing to challenge within the window, which is why the documentation describes the challenge period as a liveness requirement rather than a formality. A validity rollup's security depends on the proof system being sound, and it resolves immediately: once the proof verifies, the state is final on the base chain without a waiting period.

Both families depend on data availability, and this is the part that is easy to overlook. A rollup's state commitment is only meaningful if the data needed to reconstruct the state is actually available to anyone who wants to check it. If the data is withheld, a user cannot prove what their balance should be, and the commitment on the base chain becomes an assertion no one can verify. The Ethereum documentation treats data availability as a first-class part of the scaling design for this reason, and the base chain's own data capacity is what limits how much rollup activity it can support.

A monolithic chain asks a user to trust the chain's own consensus rules and nothing else. If the user runs a node, they verify the state themselves; if they do not, they trust that enough other people do. There is no separate operator whose honesty is required, and no window during which a result could still be reversed by a challenge.

A rollup asks for something more. An optimistic rollup asks the user to believe that someone honest is watching and will challenge a bad state root within the window, and it asks them to accept that a withdrawal may take as long as that window. A validity rollup asks the user to believe that the proof system is sound and that the verifier contract on the base chain is correct. Both ask the user to believe that the data needed to reconstruct the state is available. None of these assumptions is unreasonable, but each is an additional thing the user is relying on, and the documentation of each design states them rather than hiding them.

The comparison is not a verdict on rollups. It is a statement of what a user is buying. A rollup that posts data to a base chain with a working challenge or proof mechanism gives up some of the base chain's trustlessness in exchange for capacity, and the honest way to describe it is as a different point on the spectrum rather than as the same thing, faster.

The spectrum framing is a simplification, and it is worth saying where it breaks. Real systems sit at many points along it: a sidechain that runs its own consensus behind a federation is not a rollup, and a payment channel is not either, because its worst case is an on-chain settlement rather than a state commitment. The layer two and sidechain page draws those distinctions for Bitcoin specifically, and they apply to any network that hosts layers.

The comparison also does not measure anything. It does not say how much cheaper a rollup transaction is than a base-chain transaction, because that depends on the base chain's fee market, on how much data the rollup compresses, and on the rollup's own fee policy — all of which change. It does not say how long a withdrawal takes in practice, because that depends on the challenge window a particular rollup chooses. And it does not say which approach a given application should use, because that depends on whether the application needs the base chain's trust properties or only its settlement.

What the comparison can do is make the trade legible. Moving work off a base chain buys capacity and costs trust, and the size of both depends on the design. A reader who wants the base chain's guarantees should keep the work on the base chain and accept the capacity limit; a reader who wants the capacity should understand exactly which assumption they have taken on. Neither choice is wrong, and a page that presented one as free would be misleading.