As mentioned above under "Source", unit testing of application code modified only for year 2000 considerations is unnecessary.
You should unit test your conversion programs because they are entirely new code and embody new function.
If you can isolate the testing of data conversion and bridge programs, it is easier to isolate the cause of bugs. However, if you decided to include collateral improvements such as migration with your year 2000 changes to get the benefit of combined testing, it may not be possible to test data conversion or bridges apart from enhancements.
Under ideal conditions, the testing of data conversion and bridge programs would reveal only bugs in them. In practice, such testing uncovers latent bugs in legacy systems as well. You should track all legacy bugs whether they requiring fixing or not.
You should unit test your bridge programs because they are entirely new code or add significant new functionality to existing bridge (interface) programs.
Test wrapper style bridges via calls from application programs. For read-only bridges, testing just the bridge and receiving partitions is sufficient if the receiving partitions have been modified but the sending partition has not.
For update bridges, testing all partitions that access the data being bridged is generally required. Thus, partitions not scheduled for implementation until much later may nonetheless require some modification and testing to continue operation when other partitions go into production.
©1997 IBM Corporation