1 of 23

2 of 23

Summary

  • Rollups are in the process of decentralizing their sequencer
  • Duality of Roles: L2 sequencer = L1 proposer + L1 builder
  • We need to start separating the two roles on L2, just like we did on L1
  • PBS on L2 faces novel challenges: Privacy, Cross-domain MEV, Low latency => PBS is essential but not enough
  • We also need to decentralize the builder role itself

3 of 23

Decentralizing rollup sequencers

4 of 23

What is a sequencer?

  1. users send txns to sequencer
  2. sequencer orders them according to some policy
  3. sequencer gives some kind of receipt to the user (”pre-conf”)
  4. sequencer sends ordered batch to DA layer

Source: StarkNet

5 of 23

Why is decentralizing L2 sequencers important?

  • More and more liquidity + trading volume moving to rollups/L2s (* not the same but we’ll leave the distinction to others)
  • Today, sequencers are centralized
  • Problems with centralized sequencers:
    • Censorship
    • Liveness/availability
    • Rent extraction
    • Centralized chokepoint
    • Unattributable Frontrunning

6 of 23

How to decentralize?

  • Centralized sequencer (← we are here)
  • L1 or “other chain”-sequenced
  • Sequencer/block auction
  • Random selection from PoS set
  • FCFS (committee) sequenced

7 of 23

In L2, sequencer = L1 proposer + L1 builder

8 of 23

L2 sequencer = L1 proposer + L1 builder

If we look more closely, we can see that there are at least two components to the sequencer

  1. leader election mechanism (proposing)
  2. ordering mechanism (building)

9 of 23

How to decentralize (revisited)

10 of 23

L2s w/o PBS

Right now, these two roles are bundled together in rollup roadmaps, and this is problematic

You recreate the dynamics of pre-PBS days on Ethereum L1, leading to

  • PGAs
  • Rampant frontrunning
  • Failed txns
  • Clogged up p2p layer
  • Lower proposer revenue -> threat of vertical integration

11 of 23

12 of 23

We need to start separating the two roles on L2,

just like we did on L1.

13 of 23

Refresher: What PBS does

  • Shield proposers from the centralizing effects of MEV / make most valuable block available to all
  • Unlock competition on features from builders
  • privacy for bidders
  • more expressive for bidders (decouple position from price)
  • remove negative externalities on the chain (filter out reverts, reduce p2p load)
  • maximize revenue for the seller (of blockspace)

⇒ We need PBS + explicit MEV auctions to maximize protocol revenue and minimize negative externalities on the network + users.

14 of 23

But PBS on L2 faces novel challenges.

15 of 23

Privacy

Centralized sequencers make privacy feel easy, but decentralizing it is hard af,

16 of 23

How do we recreate privacy?

Option 1: Send to many centralized builders w centralized trust, but this is massive cognitive overhead + you enshrine some builders over others (centralizing)

Option 2: Committee-based solutions (e.g. threshold decryption, FCFS)

Option 3: Homomorphic privacy zone: super hard (what we’re working on)

17 of 23

Cross-domain MEV

Even with competition between centralized builders, you still have the problem of cross-domain MEV:

  • The builder market is naturally centralizing
  • L2s make the problem worse: the more fragmentation of domains, the more cross-domain mev. The more centralizing pressure on the builder through demand for atomicity

18 of 23

Latency

  • Rollups love their blocks to be as fast as possible but this is in natural friction with decentralization
  • We need geographical decentralization for neutrality, decentralization, robustness to regulation, fairness & more → for that, we need our systems to be insensitive to latency
  • If you lower block time too much, you encourage vertical + geographical integration between proposers + builders

https://collective.flashbots.net/t/decentralized-crypto-needs-you-to-be-a-geographical-decentralization-maxi

19 of 23

PBS is essential but not enough:

we also need to decentralize the builder role itself

20 of 23

SUAVE

Trying to do the heavy lifting on ordering for other chains:

  • Build the best MEV auction, and make it decentralized
  • Removing all elements of Flashbots/searcher/builder trust in the system
  • Privacy for freedom of expression + collaboration
  • Scalable to many domains; allow for cross-domain preferences

21 of 23

Source: https://twitter.com/jon_charb/status/1637488611745492997?s=61

22 of 23

In the rollup endgame, we’re gonna need all three:

23 of 23

Conclusion

Dear Rollups: Please make MEV a more central element of your decentralization roadmaps. Sequencer = Proposer + builder. We need to decentralize both.

  • Reach out to us to learn more about SUAVE
  • Come to our forum: collective.flashbots.net