Forget about decentralizing sequencers — Rollups need security first
A centralized sequencer “makes a lot of sense” for many rollup applications
Pla2na/Shutterstock modified by Blockworks
A recent post on X, formerly known as Twitter, by Foobar about layer-2 rollup sequencer decentralization stirred up debate in the crypto community.
He’s got a point, according to Eclipse founder Neel Somani.
Speaking to Blockworks on the Lightspeed podcast (Spotify/Apple), Somani asks, “What are you really giving up when you have a centralized sequencer?”
“You don’t give up safety or liveness, in the formal sense,” he says. “You give up liveness guarantees in the short term.”
“Usually the way that you regain liveness,” he explains, “is through some sort of forced inclusion through the [layer-1], which probably has longer block times.” But from a security perspective, Somani says, “you don’t really give up anything.”
The major trade-off, Somani says, is that the sequencer has the freedom to “re-order things as they see fit — and they can continue doing that indefinitely.”
“For some DeFi applications,” he says, “that will prove to be really difficult because you don’t want the sequencer to be trusted in doing that.” But for many apps such as games, where low latency and short confirmation times are top priorities, a centralized sequencer “makes a lot of sense,” according to Somani.
“What Foobar is saying,” adds Celestia chief operating officer Nick White, “is that rollups are supposed to be trust-minimized and ensure the state validity, or that the rules of the rollup are followed.”
“Until you have fraud proofs and ZK proofs and all that stuff,” he says, “they’re fundamentally not secure. That rollup operator could actually just steal all your money.”
Security, White argues, is far more important than the censorship-resistance and liveness that might be achieved via decentralization.
You’re always trusting something
White acknowledges that “censorship-resistance and liveness are really important features of blockchains,” pointing out that centralized sequencers could choose to reject certain transactions and are susceptible to chain-halting outages. “But I don’t put them in the same level of priority as validity and verifiability of the underlying execution.”
White mentions that a few teams, like Astria and Espresso, are working on decentralized sequencers “without having to add more overhead.” This could result in a “best of both worlds” scenario, White says, “where you get decentralization, censorship resistance and liveness guarantees, while not needing to run new infrastructure for yourself.”
“One other thing to keep in mind,” Somani says, is that “people get obsessed with the words ‘centralized’ and ‘decentralized.’”
“That’s a remnant from [layer-1s],” he says, “where that was a critical property in order for the system to operate in a trustless way, to whatever degree ‘trustlessness’ actually makes sense as a concept.”
Somani argues that it is possible for a system to contain multiple centralized components, yet remain fully verifiable and permissionless to access. “I hate using the word, ‘trustless’,” he says, “but it’s still trustless, to some degree.”
“You’re always trusting something.”
Don’t miss the next big story – join our free daily newsletter.
