CertaDNS
Skip to lesson

Prioritising · lesson 2 of 2

The ones worth skipping

After this lesson you can

Defend a decision not to deploy a control, in writing.

Assumes you have read Impact against effort, honestly.

A posture assessment that recommends every control is not an assessment. Some controls are the wrong call for a given organisation, and being able to say so in writing is what makes the rest of the recommendations credible.

The usual candidates

ControlReasonable to skip whenSay so because
BIMIThere is no registered trademark, or the brand benefit does not justify an annual certificateIt has no security property. Skipping it costs nothing defensively.
DNSSECThe team cannot operate it — no monitoring, no rehearsed key rollover, no recovery for a broken chainA broken chain is a total outage. Badly operated DNSSEC is worse than none.
DANEThe MX zone is not yours to sign, or DNSSEC is not yet boringIt fails closed. It should follow DNSSEC competence, not accompany it.
Registry lockBeyond the two or three domains whose loss would stop the businessThe friction is real and the fee is per domain.
Defensive registration at scaleBeyond the primary brand in the TLDs customers useThe space is unbounded; the budget buys an arbitrary fraction.
A monitoring serviceNobody is resourced to work the queueA queue nobody reads is indistinguishable from no monitoring, and costs money.

What is never worth skipping

Registrar account security
rua on every domain
The non-sending record set on parked domains
Removing dangling records
DMARC to at least quarantine on sending domains

All either free or the core of the thing. A report
that omits one of these has an error in it, not a
judgement call.

How to write it down

A skipped control needs three things recorded: what it would have addressed, why it is not proportionate here, and what would change the decision. The third is what turns an omission into a decision — it says the question was asked and gives the trigger to ask it again.

NOT DOING    DNSSEC on the primary domain
WOULD ADDRESS  Record tampering and cache poisoning;
             also a prerequisite for DANE
WHY NOT      No DNSSEC monitoring, no rehearsed key
             rollover, and one person who understands it.
             An expired RRSIG is a total outage.
REVISIT IF   The DNS provider offers fully managed
             signing with CDS automation, or a second
             engineer is trained.

This section is what gets read after an incident

“Not doing this, and why” is examined more closely than anything else in a report when something goes wrong. Recorded, it is a decision somebody made with reasons and a revisit trigger. Absent, it is a gap nobody can account for.

Knowledge check

An organisation has no DNSSEC monitoring, no rehearsed key rollover, and one engineer who understands it. What is the honest recommendation?

Last reviewed