This document summarizes the pandas interfaces I think we should expose to end users: https://docs.google.com/document/d/1QnvRPX_Oy5XCnow5wxBy_FNOcvaWwK_PcXxPXbIG6AE/edit?usp=sharing It shows how we can give pandas users the ability to restore to a previous version, evolve the schema, update columns, delete rows, create tables with generated columns, etc. The doc is intentionally high-level. I think it’d be good to have a high-level brainstorming exercise to discuss the “ideal end state for pandas users” without diving into the weeds with implementation details.