S3 has buckets and objects, but no native concept of a dataset. A lot of cost and governance pain traces back to that gap.