Trivial? The rows are ordered in creation order because that's how they are stored on disk. So to be able to move columns around, you either have to rewrite the

Trivial? The rows are ordered in creation order because that's how they are stored on disk. So to be able to move columns around, you either have to rewrite the whole table (needing a long-lasting exclusive lock which most of the schematic altering operations in postgres do not require) or you introduce another abstraction to keep track of the order which you then have to use for every query to get the correct position within the row you just read. This always means more complexity and a performance overhead and frankly provides next to zero value because columns are named in queries anyways (or the order becomes meaningless when joining)