The code that I have developed so far uses Win32::ODBC to do create/update a DSN and discover the driver to use, and names of the tables. DBI and DBD::ODBC are used to read each table in it's entirety and dump it to XML.

I much prefer the idea of using a fully DBI based solution to do both the schema detection work and the database reading. I will update my code and post it later today for critique and further analysis.
Thanks.