1

As far as I know, Kafka has been implemented based on the architecture of "Publish-Subscribe Pattern". I took a look at this concept, and found out that it's quite similar to "Message Queueing Pattern(Producer-Consumer Pattern)" but simultaneously has a big difference: restriction on the number of the receivers. (Publish/Subscribe vs Producer/Consumer?)

In a rigorous view, as the words "producer" and "consumer" are those used in the context of "Producer-Consumer Pattern", I believe it should have been much more clear if "publisher" and "consumer" were used instead, respectively.

Is this just a case of sacrificing rigor for the sake of convenience, or is it just that I have a leap in my logic?

Now.Zero
  • 1,147
  • 1
  • 12
  • 23

1 Answers1

0

As mentioned in the linked post, the act of subscribing to a publisher typically infers that the subscription needs to exist prior to the publishing event. A subscriber often also removes the event from the queue upon receiving the event.

Neither of these are true in Kafka since the broker is persistent and there can be zero consumers for any given topic.

Regarding a definitive answer of "why" - I suggest reaching out to the Kafka Developers mailing list

OneCricketeer
  • 179,855
  • 19
  • 132
  • 245