Technical article on LTAP
I've been watching cautiously Databricks' announcement of OLTP/OLAP unification from their summit. They dub it Lake-TAP as opposed to HTAP. Now their cofounder published this technical piece that popped up on my HN:
https://www.databricks.com/blog/lakebase-ltap-rethinking-database-storage
This made me look at it very differently from their original announcement at their conference. Basically, I think this is a great capability of their Lakebase Postgres offering (assuming it works, I still haven't had my hands on it): You store your data in Postgres but the source of truth is stored in Iceberg/Parquet on S3. That means you can now do warehousing on all the Postgres data directly because it's just our usual big data stack of Iceberg etc.
This seems like a bigger/different deal than how the company presents it. That is, it's really Postgres on Iceberg more than some grand unification theory. The fact that both are open source standards (Postgres and Iceberg) means it could have legs. Personally, I think CDC is a necessary evil for data engineering. But maybe no longer necessary?