alt Hacker NewsI don't get how this would do away with the need for some kind of ETL. Most apps use highly normalised schemas that are completely unsuitable for analytical users. Not to mention you wouldn't want to couple your app schema to your warehouse schema. Am I missing something? How would this replace traditional data warehousing?
Good point. For more complex scenarios, people would still be able to implement, for example, a Medallion Architecture to progressively improve data quality and structure. Because it is Postgres- and Iceberg-compatible (db and data), it's possible to bring more other advanced data tools when it's needed to perform data transformation and movement. Currently, we see it as a Postgres read replica for analytics. But it's easy to imagine that in the future it could be used as a standalone OSS database on top of a data lakehouse with an open format in S3.