The words this job uses

A network with no internet connectivity. Everything that enters passes through a controlled transfer, so anything that calls out simply will not work.

A chronological record of the actions performed during a change, retained to support investigation and problem resolution during and after it.

Everything affected when one system goes down — including the systems that depend on it and did not know they did. What sets a window's real risk.

A mapping of what happens to the business when a system is unavailable: who is affected, how fast, and at what cost. The basis for non-technical ranking.

The forum that approves production changes. For patching, the body that decides whether a class of window may go ahead and who carries it.

A defined period in which production changes are not permitted — quarter end, year end, a business event — except for approved emergency changes.

An update containing every fix that preceded it. Installing the latest one also closes everything that was missed before it.

A public unique identifier for one security vulnerability, in the form CVE-YYYY-NNNN. The shared language that lets tools talk about the same flaw.

An open standard scoring the technical severity of a vulnerability from 0 to 10. It measures the flaw — not the risk to the system it happens to sit on.

The Bank of Israel directive covering IT risk, information security and cyber defence. It consolidates and replaces Directives 357, 361 and 363.

A change that cannot wait for the normal process. §97 of Directive 364 requires defined procedures for assessing and approving it, and a named approver.

The date from which a product no longer receives security updates. After it, a new vulnerability will not be closed by a patch — it needs another decision.

A model estimating the probability that a vulnerability will be exploited in the near term. It complements a severity score rather than replacing it.

A catalog of vulnerabilities with evidence of exploitation in the wild. A stronger operational signal than severity, because it describes what is happening.

A defined period during which a system may be taken down or changed, agreed with its owner and known to everyone the outage affects.

Where a host sits in the network topology, from the external edge to internal management. It sets exposure, and therefore patch cadence.

How often a system gets a planned maintenance window. Usually derived from how exposed the host is and how critical the system on it is.

The work between finding a vulnerability and installing the fix: grouping, ranking, proposing a window, getting approval, and recording what happened.

The set of vulnerabilities one patch closes, together with every host it applies to. The unit a maintenance window can actually be booked for.

The second Tuesday of each month, when Microsoft publishes its monthly security rollup. The rhythm most Windows estates plan their year around.

The maximum time defined between finding a vulnerability and fixing it, by severity and asset criticality. A target you can be measured against.

What to do when the update fails: how to get back to the previous state, how long that takes, and who decides it is time to stop and revert.

How long a system may be down, and how much data may be lost. The two objectives that decide how long a maintenance window you can afford.

The person authorised to approve an outage of a system and accountable for the consequences. No window without them, and no evidence without a record.

Two windows overlapping in time on systems that depend on each other. Each was properly approved, and neither approver knew about the other.

The group a host belongs to on the internal Windows update server, which decides which approved updates it receives and in which wave.