Fancy way of obtaining technically O(n) sorting algorithms for most data types. The way this is achieved is by providing a scalable solution for doing a radix/bucket sort. Other data types are converted to bucket sortable types and compared based on a bucket sort.
The paper is pretty approachable.