* Need to do a lot more testing.  In particular the actual SQL
  engine needs vigorous testing -- I've done very simple things
  with it.

* The actual R SQLite package shares an enormous amount of code
  with both the RMySQL and ROracle drivers, so that piece should
  be somewhat stable (he says!).  On the other hand, currently
  we cannot use both drivers simultaneously (this should be an
  easy fix).

* Need to add the dbApply(), as in RMySQL.  This is very easy to do
  here (see RS_SQLite_stdCallback), but shouldn't this be better
  implemented in the RS-DBI.c to provide this service for all drivers
  that sit on top of RS-DBI.c?

* It would be nice to hook R as regular SQL functions and aggregations
  (e.g., select group1, group2, median(x) from FOO group by group1, group2;)
  Given the modularity in the SQLite C API, this is definitely not hard.
  How to hook more interesting R function that return objects is not clear
  at all, though.

* It would be interesting to play with the in-memory db version.