Daeseon Yoo

Wiki

Append-only log

An ordered, immutable sequence you only append to and read by offset. The principle under Kafka, WAL, and event sourcing — they're instances, not the idea.

·principle · distributed-systems · storage · durability

Instances

The principle

A log is an ordered sequence of records. You only ever append to the end; you never edit or reorder what's already there. Readers don't pop items off — they hold an offset and move it forward at their own pace. The log itself is the source of truth; any current "state" is a projection you rebuild by replaying the log.

That's the whole idea. Everything below is an instance of it.

How it maps to practice

When you see "what is Kafka?", the useful answer isn't a feature list — it's "it's an append-only log, here's what that buys you (replay, multiple independent readers, durability ordering) and what it costs (you manage offsets, storage grows)."

How it gets buried

"Real-time data-in-motion streaming platform." "The central nervous system of your data." "Event mesh." These describe the marketing surface of one product and hide the 50-year-old primitive (the database write-ahead log) underneath.

Say it clearly

"An append-only log: writers add to the end, each reader remembers its own position, and the log — not any snapshot — is the source of truth."

In Korean

"로그" stays as the loanword; "기록(記錄)" means a record/log in the everyday sense and loses the technical "append-only, offset-tracked sequence" meaning. "추가 전용 로그" is understood but rarely said — people just say "로그" and rely on context. The concept "이벤트를 다시 재생한다(replay)" is where Korean explanations usually click.