Excellent article and the openmetadata proejct which I lately starts to research. Being a past contributor of Datahub project, I agree your points about cons of datahub. Its system is little too complicated and the dependency of Kafka is unnecessary. It doesnt mean there is no place for the push. But people doesn't use it right anyway. I think the idea is that by push, we are supposed to make metadata's aspect more complete, such as ownerships, tags, which we normally don't have those with datasource's metastore. But developing a producer would be a different story, and in the end, no one wants to make the producer.