Hacker News new | past | comments | ask | show | jobs | submit login

My understanding (noob here) is that it allows producers to retry without fear of duplication. You still have to consider the system feeding the producer though. In Segment's example, clients might deliver their messages more than once to the API. Kafka's mechanism wouldn't detect duplicate messages sent to the producer, just that any given message a producer wants to append to Kafka won't be duplicated.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact