Layered Consensus Stack
Mathematical view of W3 consensus as a composable tower of primitives.
The consensus system is modeled as two cooperating stacks:
- Execution stack (E5 → E1) — converts workflow definitions into deterministic transcripts.
- Consensus stack (C3 → C1) — selects validators, validates attestations, and anchors evidence with ordering + finality.
Each layer owns a narrow responsibility and exposes explicit interfaces. Execution hands off transcripts; consensus feeds back membership, validation, and ordering guarantees. The separation keeps the contract tight while letting each subsystem evolve independently.
Execution Stack (E5 → E1)
Five Components
E5 · Workflow DSL — declarative YAML with triggers, actions, and verification policy.
E4 · Compiler — parses, validates references, emits bytecode + manifests.
E3 · VM / ISA — deterministic interpreter (WASM/eBPF) with gas metering.
E2 · Host Environment — explicit host calls (HTTP, storage, Smart Actions) with metered permissions.
E1 · Runtime State — deterministic replay, mempool ingestion, resource accounting.
Every validator runs this pipeline locally when it executes a workflow step. The output is a deterministic transcript (state diff + receipts) that the consensus stack can reference.
C3 · Workflow Orchestration
Runtime Protocol
Trigger consensus: deterministic validator subset decides if the workflow run begins.
Per-step loop: leader election → execute (using E5→E1) → attest.
Outputs: signed attestations referencing workflow + validator state hashes.
This protocol assumes each validator’s execution engine is deterministic. It defines who executes, how unanimous agreement is reached, and what proof artifacts must be emitted.
C2 · Protocol State
Structures Derived From Layer 1
- Workflow registry: deployed bytecode/YAML + activation bits.
Validator registry: stake, participation status, slashing history.
Attestation log: append-only record of trigger + step attestations.
Each is a deterministic state machine driven by ordered events. Implementations can be smart contracts, rollup state, or native modules—Layer 1 just guarantees ordering + finality.
C1 · Base Consensus
Primitives
Total ordering:
order(events) → sequence. All honest nodes agree on the same order.Finality:
finalize(sequence) → irreversible. Once committed, events cannot be reordered.
Examples: Bitcoin’s longest-chain rule, Ethereum’s Gasper, any BFT chain or rollup that exposes deterministic order + finality.
Information Flow
1. Register — Workflows and validators are registered on Layer 2 via events committed to Layer 1.
2. Execute — Selected validator runs the E5→E1 pipeline locally to produce deterministic results.
3. Attest — C3 orchestration reaches unanimous agreement and emits signatures referencing the execution transcript.
4. Commit — Attestations enter the C2 log, again via the ordered/finalized log of C1.