Infrastructure: make every connection accountable

Think in states, not just connections

An interface connects systems, but a dependable workflow also needs to describe what each message means. Received, accepted, partly completed, cancelled and reconciled are different states. Confusing them makes a technically successful connection operationally ambiguous.

For each handover, ask four questions: what is the input, what confirms success, who owns an exception, and what evidence remains?

A timeout does not tell the whole story

Imagine a fictional order request sent with reference A-104. The network times out before a response reaches the caller. Two explanations remain possible: the receiving system never accepted the request, or it accepted it but the reply was lost.

Blindly sending the same business request with a new reference can create a duplicate. A more informative investigation checks the original reference and recorded state before deciding what to do next. The correct retry behaviour depends on the interface contract.

In a design review, distinguish a connection failure from an unknown business outcome. They require different questions and evidence.

Connect security with operational ownership

NIST CSF 2.0 organises cybersecurity outcomes into Govern, Identify, Protect, Detect, Respond and Recover. It is a way to structure questions about responsibility and resilience, not a claim that a company has been certified.

Reference: NIST · Cybersecurity Framework FAQs

• Ownership: who approves changes and who investigates a failed handover?
• Access: what can each role read or change, and how is access withdrawn?
• Observation: how can an operator distinguish stale data from a quiet market?
• Recovery: what evidence confirms that the system and its business records agree after a restart?

Design a handover record

For the A-104 example, a compact record might include the request identifier, creation time, responsible system, last confirmed state, unresolved question and next owner. These fields make the investigation specific without exposing secret keys or passwords.

Candidates can practise explaining the difference between a technical symptom and a business consequence. Partners can use the same structure to discuss interfaces and escalation paths without assuming that a connection alone solves the workflow.

Check your understanding