FIELD NOTES

The details behind the tools.

Worked examples, practical tradeoffs, and useful limits. Read before you ship.

FIELD NOTE 01

How much randomness does an API key need?

Start with random bytes, then choose an encoding. Length and entropy are related, but they are not interchangeable.

Read field note ↗
FIELD NOTE 02

A decoded JWT is still untrusted

Readable claims are a debugging aid. They are not evidence of identity, permission or a valid signature.

Read field note ↗
FIELD NOTE 03

Cron schedules need a timezone

The same five fields can describe different instants. Check both the expression and the scheduler’s clock rules.

Read field note ↗
FIELD NOTE 04

Roll out CSP without guessing

Inventory resource origins, start with observations, then enforce a policy your application can actually satisfy.

Read field note ↗
FIELD NOTE 05

A sample is not a data contract

Schema inference gets a draft started. Business constraints need evidence beyond one example.

Read field note ↗
FIELD NOTE 06

Stage email authentication changes

Generate records offline, inventory your senders, and measure alignment before enforcing DMARC.

Read field note ↗
FIELD NOTE 07

Preview metadata without fetching a page

Compose a portable HTML fragment while keeping local images and drafts on your device.

Read field note ↗
FIELD NOTE 08

Choose an ID for the job

Randomness, ordering and compactness solve different problems. Select them deliberately.

Read field note ↗
FIELD NOTE 09

Make fixtures repeatable and clearly synthetic

A useful mock dataset should reproduce a bug without importing a real person’s information.

Read field note ↗
FIELD NOTE 10

Ignore rules are ordered rules

Combine build patterns carefully and remember that ignoring a file does not remove its history.

Read field note ↗