My page 311 paragaph 1 says something different than your page 311 paragraph 1.

Mine says: "...This order is maintained through a doubly linked list called a page chain. (Note that pages in a heap are not linked in any way to each other.) The order of pages in the page chain, and the order of rows on the data pages, is based on the definition of the clustered index. Deciding on which column(s) to cluster is an important performance consideration." (End of Paragraph)

This part isn't in my book.

"When the clustered index is created, it does in fact reorder the data physically on disk according to the cluster key, or as close to physical order as possible given the available space. After that, maintaining the clustered index is a logical operation. Rebuilding the index will again order the data physically. Reorganizing the index is an attempt to make the logical and the physical order match, without all the work of rebuilding the index...."

That being said, I don't understand why everyone is confused.

A clustered index does physically rearrange data on a disk when it is created and can continue to do so if there are page splits or when the index is rebuilt. There's no ambiguity here.

A clustered index logically tracks data on disk using doubly linked lists. This has been true for a long time.

Dropping and rebuilding a clustered index on a partition scheme may move data to other drives. The whole point of using a partition scheme is to distribute the data across multiple disks to improve performance by distributing the I/O as well as by creating logically and physically smaller partitions to search. Even if you don't use partitions, the best way to copy a table from one filegroup to another is to drop the clustered index and re-create it using a different file group. That has also been true for quite awhile.

I find it particularly amusing that some of you have the chutzpah to claim that Kalen Delaney is wrong. 🙂