So you have a second, opaque unique identifier for the frontend to fetch? At that point why not just use a single column (UUID)?
Higher density for indexes you use for joins. Also you can then use UUIDv4 to hide creation times if that's sensitive too.
Yes, this is standard. And the point is performance. The needs of user-facing ID's (short alphanumeric, randomly ordered) tend to be pretty orthogonal to the needs of internal ID's (raw bits, monotonically increasing, maybe timestamp data).
Yep, you'll probably join like 20 of those for each public uuid lookup. There are potential reasons not to do it this way, but it's a good default.
It makes more sense to have two columns yes, one autoincrement for the primary key, and another uuidv4. Because uuids are too large and makes indexes slower
So when public APIs use uuidv4 to fetch item, those uuidv4 still need to be indexed...
Right, but that's one lookup vs the many joins you're probably doing internally
This was hard enough to enforce when it was just people writing code, and once one of those stupid incrementing integer IDs go in a public API or important codepath, it can be a real pain to rip out.
This can be a valid reason. Will say though, in my experience, the AI coders have been good at respecting the comments above the PKs saying not to expose.