Introduction

The Cross-Zone Atomic Coordination Protocol (CACP) currently allows either sequencer to abort the protocol at any phase prior to Phase 3 completing, at no cost. This asymmetry enables the aborting party to avoid bearing any penalties, while the counterparty has already consumed resources for validation, signing, and state management.

This document explores mechanisms to make protocol abort costly for the aborting party. Three directions are examined in order of increasing complexity:

  1. Stake-based bond (on-chain): each sequencer deposits funds before entering the protocol; the aborting party forfeits its deposit.
  2. L2 or intermediate zone as bond coordinator: since the Logos L1 has no smart contract support, a dedicated execution environment is needed to hold and adjudicate stakes.
  3. Off-chain enforcement via optimistic fair exchange: the happy path runs entirely off-chain with no on-chain verification; the full protocol is verifiable on-chain, but only when a dispute is opened, and only for the contested step.

The on-chain approach is examined first as it is easier to reason about. The off-chain approach is derived from the same incentive structure but avoids L1 spam and third-party coordination.

Problem Statement

In the current CACP design, a sequencer that sends PROPOSE or ACCEPT and subsequently aborts incurs no direct cost. The counterparty, however, has already performed:

This cost asymmetry enables two classes of attack described in the CACP Security Analysis: resource exhaustion via repeated spurious proposals, and packing-rights manipulation via deliberate abort to stall a counterpart sequencer.

Direction 1: Stake-based Bond (On-chain)

Design Methodology

Before addressing the constraint that Logos L1 has no smart contract support, we first design an idealised version of the mechanism that assumes L1 can lock funds and enforce conditional transfers. This gives a clean, complete specification of what the mechanism must do. In Direction 3, we then examine how much of this logic can be moved off-chain, replacing L1 dependencies with cryptographic primitives.

Overview (Idealised L1 Design)

This section instantiates the running example of the Overview with concrete values. We assume for this section that the Logos L1 supports a primitive vault operation: a sequencer can lock a specified amount of tokens against a proposal identifier, with conditional release rules. No general-purpose smart contract is required — only the following operations:

Throughout this document, a proposal is identified by proposal_id = H(intent), binding all vault state to the specific cross-zone operation being coordinated.