Demonstration of a niche scenario where you want to alter a decimalised 8-bits-per-channel RGB colour and know the amount each channel should be changed by. Not very revolutionary, I just made this as a simple example. There's also no gamma correction and it is probably not possible to include without knowing the initial channels. This possibly has value in optimising code, calculating the difference every time may be unnecessary, for example, doing interpolations. dR,dG,dB are just the differences for each channel to be added/subracted to the initial color. Note that I did not include limits to the variables, you can break it. (doing so would require knowing the initial colour's channel values which is what the project is supposed to avoid. It shouldn't occur when initially calculating the difference from two colours).
