Missing pieces:

- large index update support (can read large indexes, but not update them)
  * HARD (all the knowledge exists, but lots of coding remains)
- handle more text index entry types (extended charsets, alternate encodings)
  * MEDIUM
- determine when/when not to inline memo/ole data for a row of data depending
  on how much will be written in the current row
  * MEDIUM
- implement index creation
  * VERY HARD (still missing knowledge about all the magic bits in index meta
    data)
- implement table, column, index renaming
  * EASY
- implement table, column, index deletion
  * EASY - MEDIUM