Recherche vérifiée par sources: Binance / Coinbase / OKX

Verified source facts

1. Coinbase documents that its user channel sends open orders in batches of 50 before subsequent updates, with a smaller batch indicating snapshot completion. 2. Binance documents a 60-minute listenKey lifetime, keepalive or recreation requirements and ordering guidance for the same event type. 3. OKX states that its orders channel sends no initial subscription snapshot and pushes only new orders or order updates. 4. OKX says a successful cancel or amend acknowledgement means the request was accepted, while the orders-channel update provides the authoritative result.

Comparison at a glance

1. Control test: Authentication

A trading-control disclosure is meaningful only when converted into a testable rule. For authentication, inspect private-channel credential scope. Use a sandbox, non-production context, or the smallest reversible test available, then use least-privileged read access. Capture configuration and event evidence without recording credentials, and verify the documented cancel, reject, decrement, or history behavior.

Model the failure path, not just the setup path. Ask which strategy owns the maker and taker orders, which account boundary the venue evaluates, which message reveals intervention, and how reconciliation behaves after a partial fill or disconnect. Platform documentation describes supported settings; it cannot prove that an operator configured them correctly. In this dimension, account-specific permissions is therefore an explicit audit gap.

The authentication row should end with a concrete monitoring rule tied to private-channel credential scope. Make the operator use least-privileged read access, retain the venue response, and assign an owner for exceptions. Documentation quality is a useful signal, but it does not certify the exchange or the trading system; account-specific permissions remains a separate production risk until observed.

2. Control test: Initial snapshot

For initial snapshot, inspect how open orders are bootstrapped. Use a sandbox, non-production context, or the smallest reversible test available, then define a complete snapshot boundary.

In this dimension, large-account timing is therefore an explicit audit gap.

The initial snapshot row should end with a concrete monitoring rule tied to how open orders are bootstrapped. Make the operator define a complete snapshot boundary, retain the venue response, and assign an owner for exceptions. Documentation quality is a useful signal, but it does not certify the exchange or the trading system; large-account timing remains a separate production risk until observed.

3. Control test: Incremental events

For incremental events, inspect new, amended, filled and cancelled updates. Use a sandbox, non-production context, or the smallest reversible test available, then map every event to one state transition.

In this dimension, undocumented edge cases is therefore an explicit audit gap.

The incremental events row should end with a concrete monitoring rule tied to new, amended, filled and cancelled updates. Make the operator map every event to one state transition, retain the venue response, and assign an owner for exceptions. Documentation quality is a useful signal, but it does not certify the exchange or the trading system; undocumented edge cases remains a separate production risk until observed.

4. Control test: Connection lifetime

For connection lifetime, inspect listenKey or session expiry. Use a sandbox, non-production context, or the smallest reversible test available, then renew before expiry and alert on failure.

In this dimension, network interruption is therefore an explicit audit gap.

The connection lifetime row should end with a concrete monitoring rule tied to listenKey or session expiry. Make the operator renew before expiry and alert on failure, retain the venue response, and assign an owner for exceptions. Documentation quality is a useful signal, but it does not certify the exchange or the trading system; network interruption remains a separate production risk until observed.

5. Control test: Ordering

For ordering, inspect sequence, event time and transaction time. Use a sandbox, non-production context, or the smallest reversible test available, then reject impossible state regression.

In this dimension, cross-type ordering is therefore an explicit audit gap.

The ordering row should end with a concrete monitoring rule tied to sequence, event time and transaction time. Make the operator reject impossible state regression, retain the venue response, and assign an owner for exceptions. Documentation quality is a useful signal, but it does not certify the exchange or the trading system; cross-type ordering remains a separate production risk until observed.

6. Control test: Deduplication

For deduplication, inspect repeated event or fill identifiers. Use a sandbox, non-production context, or the smallest reversible test available, then make processing idempotent.

In this dimension, provider-specific duplicates is therefore an explicit audit gap.

The deduplication row should end with a concrete monitoring rule tied to repeated event or fill identifiers. Make the operator make processing idempotent, retain the venue response, and assign an owner for exceptions. Documentation quality is a useful signal, but it does not certify the exchange or the trading system; provider-specific duplicates remains a separate production risk until observed.

7. Control test: Acknowledgement

For acknowledgement, inspect request accepted versus final result. Use a sandbox, non-production context, or the smallest reversible test available, then wait for authoritative order update.

In this dimension, delayed terminal event is therefore an explicit audit gap.

The acknowledgement row should end with a concrete monitoring rule tied to request accepted versus final result. Make the operator wait for authoritative order update, retain the venue response, and assign an owner for exceptions. Documentation quality is a useful signal, but it does not certify the exchange or the trading system; delayed terminal event remains a separate production risk until observed.

8. Control test: Reconnect

For reconnect, inspect gap after disconnection. Use a sandbox, non-production context, or the smallest reversible test available, then refresh REST state before replay.

In this dimension, events lost during gap is therefore an explicit audit gap.

The reconnect row should end with a concrete monitoring rule tied to gap after disconnection. Make the operator refresh REST state before replay, retain the venue response, and assign an owner for exceptions. Documentation quality is a useful signal, but it does not certify the exchange or the trading system; events lost during gap remains a separate production risk until observed.

9. Control test: Reconciliation

For reconciliation, inspect stream state versus authoritative query. Use a sandbox, non-production context, or the smallest reversible test available, then run periodic and post-error checks.

In this dimension, eventual consistency is therefore an explicit audit gap.

The reconciliation row should end with a concrete monitoring rule tied to stream state versus authoritative query. Make the operator run periodic and post-error checks, retain the venue response, and assign an owner for exceptions. Documentation quality is a useful signal, but it does not certify the exchange or the trading system; eventual consistency remains a separate production risk until observed.

10. Control test: Audit evidence

For audit evidence, inspect non-secret raw event and decision trail. Use a sandbox, non-production context, or the smallest reversible test available, then retain IDs, timestamps and state transitions.

In this dimension, retention policy is therefore an explicit audit gap.

The audit evidence row should end with a concrete monitoring rule tied to non-secret raw event and decision trail. Make the operator retain IDs, timestamps and state transitions, retain the venue response, and assign an owner for exceptions. Documentation quality is a useful signal, but it does not certify the exchange or the trading system; retention policy remains a separate production risk until observed.

Decision scenarios

Scenario 1: when authentication becomes decisive

Consider a reader whose main constraint is private-channel credential scope. The correct response is not to infer a universal platform or product winner. It is to use least-privileged read access, retain the evidence, and define a stop condition before funds or permissions are committed. If the evidence changes, the decision must be reopened. This produces a repeatable operating rule rather than a one-time opinion.

Scenario 2: when initial snapshot becomes decisive

Consider a reader whose main constraint is how open orders are bootstrapped. It is to define a complete snapshot boundary, retain the evidence, and define a stop condition before funds or permissions are committed.

Scenario 3: when incremental events becomes decisive

Consider a reader whose main constraint is new, amended, filled and cancelled updates. It is to map every event to one state transition, retain the evidence, and define a stop condition before funds or permissions are committed.

Scenario 4: when connection lifetime becomes decisive

Consider a reader whose main constraint is listenKey or session expiry. It is to renew before expiry and alert on failure, retain the evidence, and define a stop condition before funds or permissions are committed.

Scenario 5: when ordering becomes decisive

Consider a reader whose main constraint is sequence, event time and transaction time. It is to reject impossible state regression, retain the evidence, and define a stop condition before funds or permissions are committed.

Scenario 6: when deduplication becomes decisive

Consider a reader whose main constraint is repeated event or fill identifiers. It is to make processing idempotent, retain the evidence, and define a stop condition before funds or permissions are committed.