On-chain settlement infrastructure

The situation

A client needed settlement between parties who did not share a ledger, with finality and an audit trail. The question we were asked first was not how to build it but whether it needed a chain at all — which is the right question, and in this case the answer was yes, because no party was willing to hold the authoritative record.

What we built

Settlement contracts written to be audited, with the failure paths designed first: partial settlement, stuck transactions and reorgs are normal operating conditions in this environment, not exceptions to handle later. Around the contracts sits the part that is ordinary distributed-systems engineering — indexing, reconciliation, monitoring and alerting.

• Settlement contracts with full test coverage, prepared for third-party audit
• Off-chain indexing so settlement history is queryable like any other data
• Reconciliation between on-chain state and the client's internal records
• Monitoring and alerting on stuck or failed settlements

Our broader view on where distributed ledgers belong — and where they do not — is in our notes from the Istanbul conference.