Rebuilding databases where the original database collation does not match the character set stored in the database can be tricky. A common issue in early SA databases was databases using collations with cp850 (DOS character set) talking to Windows applications loading the database with cp1252 (Windows character set). Since early databases did not have char set conversion, or had it but it was off by default, it was possible to work for years without realizing that the data and collation did not match.

To fix a situation like this, the essential steps are:

1. Identify exactly what char set is in the old database.
2. If the char set and collation don't match, unload the database without char set conversion. This ensures that the data you get in your output files has not been mangled by conversion. The data is in the char set that was stored in the database. If possible, use an editor that can handle the char to view the data and ensure that it looks correct.
3. Create your new database with a compatible collation, one that can handle all of the characters in the data to be loaded.
4. Load the new database, specifying the char set of the unloaded data. Char set conversion should be enabled assuming that you have correctly identified the unloaded data's char set and have selected a compatible collation.

If the original database contains data with mixed and incompatible char sets, manual cleanup will be required. Unload the data without conversion, use an editor that can handle the characters in the files, and clean up the data before attempting to load it into the new database.

Here is a description of some of the considerations and the tools available in SA12: http://dcx.sybase.com/index.html#1201/en/dbadmin/changing-collation-natlang.html*d5e29036