Understanding the Layer 3 appchain landscape on Starknet

🤷‍♀️ All-Purpose Hangout
1

Complete noob in the Starknet layer 3 landscape here to ask questions to help me better understand it. :thought_balloon:

As of my current understanding, the collaboration with Celestia is a major step towards providing a data availability layer for L3s on Starknet.

The vision (please correct me if I’m missing something) is that Starknet processes proofs submitted by L3s (with the help of sequencers like Madara), each L3 being a specialized blockchain that could run its own consensus mechanism, tokenomics and specialize in its area of expertise. Currently Dojo is looking at this landscape to enable project teams to deploy games as L3s, reducing costs and making it more scalable. This opens new doors to the builder journey on Starknet whereby it’s now possible to deploy a dapp as a sovereign state machine secured by Ethereum.

Highlevel questions:

  • What key architectural components are required on the L3 side to have a Starknet-compitible L3? Does it need to be running a Cairo VM?

I’m curious to double click into what this builder journey looks like:

  • What existing frameworks will be used to build an L3?
  • How easy will it be to deploy an appchain? What does the L3 onboarding mechanism look like? Is it completely permissionless?
  • What overhead and tradeoffs exists with maintaining your own L3 vs. deploying contracts directly to Starknet?
  • Will these L3s have their own on-chain governance mechanisms?
  • Is it possible to have two different L3s send messages between each other? How?

Madara will also have a crucial role by providing sequencers for these L3s.

  • Will Madara have specialized sequencers dedicated to serve specific L3s?
  • What do the economics look like for L3s that require Madara to maintain uptime? And for Madara to “service” them?
2

Hi, I’ll answer some of these q’s the best I can

  1. What key architectural components are required on the L3 side to have a Starknet-compitible L3? Does it need to be running a Cairo VM?
    The same way that Starknet does not need to be EVM-compatible in order to use Ethereum for settlement/DA/consensus, L3s do not need to be Cairo-compatible. An L3 handles its own execution, so that execution does not need to be compatible with any env on a layer beneath it

  2. What existing frameworks will be used to build an L3?
    I imagine StarkEx will be used, which can allow L3s to share proving with Starknet

  3. How easy will it be to deploy an appchain? What does the L3 onboarding mechanism look like? Is it completely permissionless?
    It’s completely permissionless. Not sure about the level of difficulty

  4. What overhead and tradeoffs exists with maintaining your own L3 vs. deploying contracts directly to Starknet?
    The overhead is that you as the L3 operator must run nodes (sequencer, prover or shared proving, etc) and some infra like RPCs and explorers. What you gain from this work is a fully customizable execution environment and dedicated throughput

  5. Will these L3s have their own on-chain governance mechanisms?
    Depends on what the team wants to do, but I would expect most to do this

  6. Is it possible to have two different L3s send messages between each other? How?
    Absolutely. They can bridge to each other using a 3rd party, or first touch down to Starknet. But if they want to communicate without a step in between, they will likely need a shared sequencer (eg they can both use Madara). Espresso is a shared sequencer for eth L2s which can give you some great insight on how this process can work and what you gain from it