L1 / L2 / off-chain interactions

Myelin is not one layer — it's a bridge across three layers. This section is the visual map of those layers and the protocols that move work between them.

What "L1", "L2", "off-chain" mean here

For Myelin, the three layers mean very specific things:

L1 CKB mainnet / testnet / devnet. The Nervos CKB proof-of-work chain. Where the long-lived asset custody lives, where the future court verifier lives, and where any "submitted" piece of evidence ends up.

L2 Myelin session runtime. A bounded finite-Cell ledger with deterministic CKB-VM-style execution, a CellDAG scheduler, and a selectable finality engine. Where high-throughput transitions happen.

Off-chain Producers, witnesses, and data availability. The producers that submit CellTxs, the witnesses (e.g. game tape, signed quotes, sensor batches), and the DA store / external DA provider that holds the chunk payloads.

%%{init: {
  "theme": "base",
  "themeVariables": {
    "primaryColor": "#A5B4FC",
    "primaryTextColor": "#1E293B",
    "primaryBorderColor": "#4F46E5",
    "lineColor": "#6366F1",
    "secondaryColor": "#C7D2FE",
    "tertiaryColor": "#C7D2FE"
  }
}}%%
flowchart TB
    subgraph OFF["Off-chain"]
        P["Producer<br/>(game server / IoT gateway /<br/>market maker / agent)"]:::off
        W["Witness<br/>(game tape / signed quote /<br/>sensor batch)"]:::off
        D["DA store<br/>+ external DA receipt"]:::off
    end
    subgraph L2["L2 — Myelin session runtime"]
        M["Mempool<br/>admission queue"]:::l2
        S["CellDAG<br/>scheduler"]:::l2
        V["CKB-VM-style<br/>verifier"]:::l2
        C["Committee<br/>certificate<br/>(static / Tendermint)"]:::l2
    end
    subgraph L1["L1 — CKB"]
        A["Asset custody<br/>(session lock Cells)"]:::l1
        E["DA anchor<br/>CellTx package"]:::l1
        S2["Settlement<br/>CellTx package"]:::l1
        CT["Court verifier<br/>(future)"]:::l1
    end

    P --> W --> M
    M --> S --> V --> C
    V --> D
    C --> E
    C --> S2
    A -.->|lock at open| M
    A -.->|settle at close| S2
    E --> CT
    S2 --> CT

    classDef off fill:#CBD5E1,stroke:#334155,color:#1E293B;
    classDef l2  fill:#A5B4FC,stroke:#4F46E5,color:#1E293B;
    classDef l1  fill:#C7D2FE,stroke:#7C3AED,color:#1E293B;

The dotted arrows are long-lived custody flows (session open and session close). The solid arrows are evidence flows (CellTx, execution, projection, DA, submission).

Pages in this section

• The three-layer model

  ---

  The full picture: who owns what, who computes what, who can dispute what, and what the future court path looks like.

• Session lifecycle

  ---

  Open → commit → DA → court → settle, step by step, with the artefacts at each step.

• Court path

  ---

  What happens when a chunk is disputed: bundle construction, bundle verification, and the future CKB-VM court verifier.

• Data availability flow

  ---

  DA manifest, segment proof, local store, external receipt, anchor package, and the readiness ladder.

• L1 submission flow

  ---

  From a verified evidence package to a CKB RPC submission, with the context / economics / inclusion / stability / finality readiness chain.

Reading order

If you're new to Myelin:

1. The three-layer model — get the picture.
2. Session lifecycle — see one full cycle.
3. The other three pages are deep dives by topic.

If you're auditing Myelin against another L2:

1. The three-layer model — compare to your model.
2. Court path — read this carefully. The court shape is what makes Myelin "CKB-aligned" rather than "yet another sidechain."
3. Data availability flow and L1 submission flow — the rest of the credibility hinges.