Databases have evolved to efficiently separate data storage from data access, with many modern databases not even containing any actual data. DuckDB is a prime example of this evolution, allowing users to create virtual datasets that can be accessed without downloading the entire dataset. By creating a database file with specific instructions on how to interact with the data, users can easily share and analyze complex datasets without the need for massive downloads or complicated data transfer processes. DuckDB’s ability to support partial reading and dynamic data changes ensures that users can access relevant data with ease, making relational datasets just a hyperlink away.
https://www.nikolasgoebel.com/2024/05/28/duckdb-doesnt-need-data.html