Interoperability: The Quiet Revolution
Six months into working with UniForm and the broader interoperability story from DAIS 2024, and I want to document what's actually happening in production. The interoperability revolution that was announced at Summit is real — it's also slower and more complicated than conference demos suggest.
What's Actually Working
Delta-to-Iceberg reads via UniForm are production-ready for the use case they were designed for: allowing Iceberg-compatible engines to read Delta tables without requiring data export or format conversion. I've validated this with Athena and with a customer running Starburst. The reads work. The column statistics are used for predicate pushdown. Performance is comparable to native Delta reads from those engines.
The caveat that matters: this is read-only interoperability. The Iceberg-compatible engine reads the Delta table; it doesn't write to it. Bidirectional writes are not supported, and won't be in the near term. If you need to write from multiple engines to the same table, you still need to pick one engine's native format as the write format and use the other as a read-only consumer.
Why Open Formats Will Define the Next Decade
The data platform landscape is not going to consolidate into one vendor. Every large enterprise I work with runs data across multiple clouds, in multiple tools, with multiple teams that made independent platform choices over the last ten years. The realistic trajectory is not "one platform to rule them all" — it's "a set of open formats that all platforms can read, governed by a catalog that all platforms can query."
This is the future that UniForm, Delta Sharing, and open-source Unity Catalog are building toward. The table format is an open spec. The catalog is an open spec. The sharing protocol is an open spec. Any compute engine that implements these specs can participate in the ecosystem. This is the right architecture for a multi-vendor world.
What Vendors Will Do Next
The vendors that are farthest behind on interoperability are going to be forced to catch up or lose deals. The conversation with enterprise architects is increasingly "can your platform read from and write to the open formats we're standardizing on?" Vendors who answer "not natively, but we have an export tool" are going to lose to vendors who answer "yes, we're UC-compatible and we read Delta natively."
Expect Snowflake to deepen their Iceberg native support. Expect dbt to extend its catalog integration with Unity Catalog. Expect the major cloud providers to build tighter Unity Catalog integrations as the spec stabilizes. The format wars incentivized fragmentation. The interoperability layer is now incentivizing convergence. As always, I'm here to help.