Adam wrote:

>
> Bidirectional indices take up more space and so must be less dense
> than unidirectional indices. Having larger indices means that you have
> more sparse index pages and reading the index to will take longer.

The real problem is that bi-directional indexes lead cause internal
deadlocks unless one direction or the other is treated as unsafe.
The code to handle unsafe pointers is large, slow, and tends to be
buggy.

Regards,


Ann