Example: a well-structured Markdown page

Information type:Concept•"What is it?"

Understanding, background, definitions, and mental models

Worked example

This is a finished illustration to study, not a template. For the blank structure, see the concept article template. To learn about the concept information type in depth, see Information types.

The source

---
title: "Idempotency"
description: "What it means for an operation to be idempotent, and why it matters for retry logic in distributed systems."
contentType: concept
---

## Definition

An operation is **idempotent** if performing it multiple times has the
same effect as performing it once. `PUT /users/42 {"name": "Alice"}`
is idempotent — running it five times leaves the same end state as
running it once. `POST /users {"name": "Alice"}`, which creates a new
user each time, is not.

## Why it matters

Networks fail. A client that doesn't receive a response can't tell
whether the request succeeded before the connection dropped. If the
operation is idempotent, the client can safely retry. If it isn't, a
retry risks a duplicate side effect — a second user record, a second
charge.

## Key principles

- Idempotency is a property of the *operation*, not the transport —
  retrying at the HTTP layer doesn't make a non-idempotent operation
  safe.
- `GET`, `PUT`, and `DELETE` are idempotent by HTTP convention;
  `POST` and `PATCH` generally are not, unless the API explicitly
  guarantees it (often via a client-supplied idempotency key).
- An idempotency key lets a client make an inherently non-idempotent
  operation (like `POST`) safe to retry, by having the server
  deduplicate requests that carry the same key.

## Examples

Safe to retry without a key:

```http
PUT /users/42
{"name": "Alice"}
```

Not safe to retry without a key — a dropped response could result in
two users named Alice:

```http
POST /users
{"name": "Alice"}
```

## Trade-offs

Idempotency keys add server-side state (the server must remember
which keys it has already processed, for some retention window) and
a small amount of client complexity (generating and storing the key).
For low-stakes, rarely-retried operations, this cost may not be worth
it.

## Related concepts

- [Retry logic](#)
- [At-least-once vs. exactly-once delivery](#)

Why this page is structured this way

• The description does real work. It’s specific enough to tell a search result apart from a dozen other API articles, and short enough to read in one glance.
• Definition leads with the positive example, then the negative one - showing the boundary of the concept is often clearer than defining it in the abstract.
• Why it matters comes before Key principles. A reader who doesn’t yet care about idempotency won’t get far enough to learn the principles; motivate first.
• Trade-offs is genuinely earned here - idempotency keys have a real cost (server-side state), so the section isn’t padding. Compare this to a concept with no real trade-off, where the template says to drop the section entirely.
• Related concepts links to real adjacent ideas, not a generic “see also” - each link should be something the reader plausibly needs next.