This page discusses database primary key ID technologies, tracing their historical evolution and presenting current best practices. It compares various ID schemes including AUTO_INCREMENT, UUIDv4, Snowflake, ULID, CUID2, and UUIDv7 across criteria such as time sorting, distributed generation, guessing resistance, and database compatibility. The article explains limitations of sequential IDs—specifically single points of failure in sharding and security risks from predictability. UUIDv7 emerges as the modern optimal solution, offering time sorting, distributed generation, guessing resistance, and database compatibility. The recommended approach is adopting UUIDv7 for internal databases while using Base62 encoding or NanoID for externally exposed identifiers.
