>What would be the best way to manage a possible mess?

Denis,

We have an application which always uses one single set of source code (producing a single executable). When clients ask for custom work, we usually document the specs and have them sign off on the time/cost estimates for the work.

At coding time, we usually set up #Defines to represent the new custom functionality based on the licensee code, since we are aware of the licensee via the data. Yes, that could make for a few IF statements here and there to only process certain code for certain clients. However, it keeps us from having to maintain separate code sets for different clients. As Kevin said, that is a huge headache.

The other thing I'll mention is that some of the custom work is represented in our class hierarchy, as opposed to IF statements. For example, a lot of our clients interface our data to their general ledger system. Each client's interface format would obviously be different. So, we created a parent GL interface class and each client's implementation is a subclass. Subclassing is a very clean way to handle custom work.

Hope that was helpful.

Laterness,
Jon