I've eliminated some errors, and found how to convert the international characters. If you are indeed using international characters (for instance the é in Méndez), you will need to install python-core and python-netclient. If you are not using any special characters, or if you're not sure, just try running the script. It will run fine without python unless it detects some base64 encoding.
This resolves all except for one issue... If you copy an ldif file directly over from Windows, it will have linefeeds at the end of each line. Line 39 of the script is currently stripping the last character, however it does not do a proper test for the linefeeds. If I can find how to get SED to strip these, then all other parsing of the ldif files appears to be handled correctly.
Note that I still have not added the full list of fields to the conversion.
Updated script is available at
http://sourpuss.net/ldif2xml/[EDIT] I found a solution to stripping the CrLf by using the TR command. New script uploaded, so it should work flawlessly except for the missing fields.