Yeah, rowid is SQLite-specific, but most^{[citation needed]} major database engines will generate some kind of hidden unique identifier for each row if there isn't one explicitly specified. I believe this is because it's much harder to implement joins between tables which completely lack any way to uniquely identify rows. (I read something along those lines many years ago, but haven't really looked into it in any detail.) IIRC, MySQL and Postgres use UUIDs for this purpose and, of course, the field names involved will be engine-specific (if they're user-accessible at all).