Suppose that you have a large project supported by an API base. The project also ships a public API that end(ish) users can use.
Sometimes you need to make changes to the API base that supports your project. For example, you need to add a feature that needs an API change, a new method, or requires altering of one of the objects, or the format of one of those objects, passed to or from the API.
Assuming that you are also using these objects in your public API, the public objects will also change any time you do this, which is undesirable as your clients may rely on the API objects remaining identical for their parsing code to work. (cough C++ WSDL clients...)
So one potential solution is to version the API. But when we say "version" the API, it sounds like this also must mean to version the API objects as well as well as providing duplicate method calls for each changed method signature. So I would then have a plain old clr object for each version of my api, which again seems undesirable. And even if I do this, I surely won't be building each object from scratch as that would end up with vast amounts of duplicated code. Rather, the API is likely to extend the private objects we are using for our base API, but then we run into the same problem because added properties would also be available in the public API when they are not supposed to be.
So what is some sanity that is usually applied to this situation? I know many public services such as Git for Windows maintains a versioned API, but I'm having trouble imagining an architecture that supports this without vast amounts of duplicate code covering the various versioned methods and input/output objects.
I'm aware that processes such as semantic versioning attempt to put some sanity on when public API breaks should occur. The problem is more that it seems like many or most changes require breaking the public API if the objects aren't more separated, but I don't see a good way to do that without duplicating code.