Hacker News Reader

I’m still waiting for Jepsen to put Confluent’s “Kafka provides exactly once delivery semantics” claim to the test.

Since they’re claiming something provably false, it’d be nice to have some empirical evidence as such.

I'm not convinced it is false--IIRC their claim is specifically w.r.t other Kafka side effects, and those they can control.

I thought they only guaranteed exactly once processing ? That could just be idempotency.