When programming an application, one should first design the (G)UI and only then write the application code itself. If you write the code first, you might end up not being able to shape the UI in a certain way without refactoring or otherwise changing the code.

Similarly, when designing an application that uses a proprietary protocol or file format, first design the UI. Then, model the protocol or file format based on the correct way to implement it, given the already designed UI. This way, the chance to have missing or redundant features in the protocol is smaller, and less time's wasted.

More features as well as compatibility, generality, scalability, flexibility and other "bilities" will follow.

I usually find that I need to add some overlooked option to my UI after I have written the code, so for me the reverse seems to often be true.  So far I haven't been unable to shape a UI around any code, though, so my experiences might not be applicable.

I'm not all that big on object-oriented programming at the moment. I write rock-paper-scissors games in Java. =)

Hm, I deleted my last post, since my programming has changed (for the better!) I have realized that there is a lot more leniency in programming than what my last post described :P