-1. Clean up the C code a lot. Much work is duplicated and a bunch of
malloc/free calls could be avoided. There's plenty room for
optimization. 

0. Write a Python layer over the C layer to allow easier
customization. 

1. Better support for LONG VARBINAR/VARCHAR variables. Perhaps accept
a Python file object as variable.

2. MS cursor library emulation to allow scrolling a cursor
backward. Currently this as to be managed by the application.

3. Interface to various ODBC options (Transaction isolation level, statement
options, ...).

4. Write a hfetchone(), hfetchmany() method which doesn't return a row
as a tuple but as a dictionary, so you can access the columns by name
instead of position. Can be don in Python (see item 0.)

5. Better exception mechanism. As soon as i know how to import a
Python class into a C module and instantiate Python objects and use
these objects as exceptions, the OSBC error message can be split in
it's part and the applikcation doesn't have to do string parsing to
decide if the error was fatal or the action should be rertried.

6. Write a Tkinter Interface to the database for administration,
testing purposes.

7. Write documentation, examples, applications, ...

7. Misc other things.
