Sheer nitpick here...
A B-tree is where the records (data) live at all levels of the tree;
B+ tree is where the records are only at the leaf level.
That's what Knuth calls them, anyway.
Clustered indexes for all known dbs are true B+ trees.
Nonclustered indexes could be B-trees (probably aren't),
since there's no big fanout penalty for storing the little
(heap) row locators everywhere at all levels.
J. Andrew Rogers wrote:
> As far as I know, Oracle does it by having a B-Tree organized heap (a
> feature introduced around v8 IIRC), basically making the primary key
> index and the heap the same physical structure.
...
