How is a spatial index represented

I read the documentation at msdn abou开发者_如何转开发t spatial indexes but I do not understand two things,

• what are the keys for the index.
• how and where are the sets of covered cells represented in the B-tree.

---

The keys for the index should be polygons (in this case rectangles) in the space that is being indexed (or equivalent to i.e. they probably just use the grid index), in this case these appear to be on a regular, hierarchical grid.

As stated here a variation of the hilbert space filling curve is used to map the 2d index to a 1d B-tree. This is essentially to provide good spatial locality between close indexes.