All objects we create, must be added manually to source control, and this is by design. The reason for this is that we may not be at a stage where the code, which we're working on, is ready for other developers. It could also be that we have various code snippets that are just for our personal use, and we don't want them published to the team.

You can add a new object to version control as soon as it is created, but we delay checking in the code until we are at least approaching code complete for a section of the work. This is because checked-in code is available to other developers (and the build process) when they choose to synchronize.

In some cases, we will be writing foundation code and data structures that are required by other team members; internal communication is therefore important.

We tend to use a discrete project for each packet of work, which is owned by one developer. The code is added to source control once the basic framework of objects is complete, but is not checked in until it is code complete. If an element needs to be renamed, it has to be checked in first, so don't add it until you are sure your objects (such as classes and tables) don't need to be refactored. It may be argued that a good technical design should avoid this scenario, but it can still happen.