1. An organization wants to share a Delta table containing aggregated sales data with a partner company that uses a non-Databricks platform (e.g., Apache Spark on their own infrastructure). The data must remain in the provider's storage and the partner should only have read access. Which Databricks feature is MOST appropriate for this use case?
- A. Unity Catalog external location shared via IAM cross-account role
- B. Delta Sharing, using an open protocol that allows the recipient to access data without a Databricks account✓ Correct
- C. Unity Catalog with cross-workspace catalog binding to the partner's workspace
- D. Databricks Repos with the Delta table exported as Parquet files and pushed to a shared Git repository
Explanation
Delta Sharing is an open protocol specifically designed to share live Delta tables with external parties, including those not using Databricks. The data stays in the provider's storage, and recipients receive read-only access via a sharing URL and credential token. Option A is incorrect because sharing an IAM role with an external party is a security anti-pattern and does not provide the governed, read-only sharing model needed. Option C is incorrect because cross-workspace catalog binding is for sharing within the same Databricks organization/metastore, not with external non-Databricks parties. Option D is an anti-pattern that involves copying data, breaks the 'live' requirement, and introduces security and governance risks.