Question about Kafka and zookeeper

Issue

• I've been reviewing documentation on clients connecting to zookeeper and am finding some mixed responses. It looks like in the most recent versions of kafka neither producers or consumers need to connect directly to zookeeper and can connect directly to one of the kafka nodes (using a list of bootstrap servers).

• In the past, information about offsets was stored only in zookeeper - which is why consuming applications used to connect there.

• So, moving forward, should all clients connect to kafka? Or is there some other reduction in functionality in bypassing zookeeper?