FAQs: Protocol

What's the difference between Arbitrum Rollup and Arbitrum AnyTrust?

Arbitrum Rollup is an Optimistic Rollup protocol; it is trustless and permissionless. Part of how these properties are achieved is by requiring all chain data to be posted on layer 1. This means the availability of this data follows directly from the security properties of Ethereum itself, and, in turn, that any party can participate in validating the chain and ensuring its safety.

By contrast, Arbitrum AnyTrust introduces a trust assumption in exchange for lower fees; data availability is managed by a Data Availability Committee (DAC), a fixed, permissioned set of entities. We introduce some threshold, K, with the assumption that at least K members of the committee are honest. For simplicity, we'll hereby assume a committee of size 20 and a K value of 2:

If 19 out of the 20 committee members and the Sequencer are malicious and colluding together, they can break the chain's safety (and, e.g., steal users' funds); this is the new trust assumption.

If anywhere between 2 and 18 of the committee members are well behaved, the AnyTrust chain operates in "Rollup mode"; i.e., data gets posted on L1.

In what should be the common and happy case, however, in which at least 19 of the 20 committee members are well behaved, the system operates without posting the L2 chain's data on L1, and thus, users pay significantly lower fees. This is the core upside of AnyTrust chains over rollups.

Variants of the AnyTrust model in which the new trust assumption is minimized are under consideration; stay tuned.

For more, see Inside AnyTrust.

Are “Sequencers” the same thing as “Validators”? Can a centralized Sequencer do bad things like steal all my money?

No and no!

An Arbitrum Chain's Sequencer(s) and Validators and completely distinct entities, with their own distinct roles.

The Sequencer is the entity granted specific privileges over ordering transactions; once the Sequencer commits to an ordering (by posting a batch on Ethereum), it has no say over what happens next (i.e., execution). A malicious/faulty Sequencer can do things like reorder transactions or temporarily delay a transaction's inclusion — things which could be, to be sure, annoying and bad — but can do nothing to compromise the chain's safety.

The Validators are the ones responsible for the safety of the chain; i.e., making staked claims about the chain state, disputing each other, etc.

Currently, on Arbitrum One, the Sequencer is a centralized entity maintained by Offchain Labs. Eventually, we expect the single Sequencer to be replaced by a distributed committee of Sequencers who come to consensus on transaction ordering. This upgrade will be an improvement; we don't want you to have to trust us not to reorder your transactions. However, it also isn't strictly necessary for Arbitrum One to achieve its most fundamental properties.

In other words: An Arbitrum Rollup chain with a centralized Sequencer could theoretically still be trustless!

Which is to say — the more important thing than decentralizing the Sequencer, i.e., the thing you ought to care more about — is decentralizing the Validators.

Arbitrum One's Validator set is currently whitelisted; over time, we expect governance to expand the whitelist and eventually be removed entirely. For more info see "State of Progressive Decentralization".

If there is a dispute, can my L2 transaction get reorged / throw out / "yeeted"?

Nope; once an Arbitrum transaction is included on L1, there is no way it can be reorged (unless the L1 itself reorgs, of course). A "dispute" involves Validators disagreeing over execution, i.e., the outputted state of a chain. The inputs, however, can't be disputed; they are determined by the Inbox on L1. (See Transaction Lifecycle)

...okay but if there's a dispute, will my transaction get delayed?

The only thing that a dispute can add delay to is the confirmation of L2-to-L1 messages. All other transactions continue to be processed, even while a dispute is still undergoing. (Additionally: in practice, most L2-to-L1 messages represent withdrawals of fungible assets; these can be trustlessly completed even during a dispute via trustless fast "liquidity exit" applications. See L2-to-L1 Messages).

Why was “one week” chosen for Arbitrum One’s dispute window?

Generally, some dispute period is necessary for Arbitrum Rollup so that validators have time to dispute an invalid assertion.

A week is expected to be more than enough time for validators to carry out an interactive dispute, assuming they don’t encounter difficulty in getting their transactions included on L1. One week was chosen following the general consensus among the Ethereum research community — as well as other layer 2 projects — to provide enough time for the community to socially coordinate in the case of a coordinated Ethereum-staker censorship attack.