Comment Re:What does this mean? (Score 1) 20
All the functional checks are done in the producer and consumer client code - the only thing any Confluent hosted tier does is check to see whether the schema-encoded Kafka message contains a schema ID that matches one for that topic, it does absolutely no data validation otherwise.
So, if you have a bad client, you can publish data to a topic which does not validate against any schema, but the topic will accept it so long as the schema ID presented is valid. The entire thing is based on trust.
You can do much better validation than their implementation, essentially, and lose nothing.