Evolution vs Design in Software

I am finally nearing delivery of my largest project to date at my day job, so it seems appropriate to share some of what I learned during its course. The topic today is that of software design vs software evolution, a fine balance which I didn't get totally right with the application we were building.

By software design, I essentially mean planning the high level structure of the section of the codebase in question - likely by writing something akin to a software design description, although in our case it would've been significantly less formal. In all likelihood, we would've just written down a bunch of classes and methods and wrote down what each one would need to do to get them to work. By software evolution, I mean eschewing that step, and simply mashing away at your keyboard with reckless abandon, with your code evolving into something with the functionality you want over time.

The former option might sound like a no-brainer, but I think there are two important points to note in favour of the latter, particularly from the point of view of a relatively inexperienced developer such as myself. Firstly, when you are writing your SDD, there are things that you don't know you don't know - so it is difficult to account for them in your skeleton code. There are inevitably going to be challenges that you didn't foresee that mean you have to make changes to your SDD, and perhaps your code architecture has to change as a result. Secondly, I am a big believer in the 'all code is garbage' mantra. With the evolutionary approach, you acknowledge and even expect to have to rewrite most of your code. Your first attempt is essentially a reconnaissance mission, to be replaced by the full army of your programming prowess once the lay of the land has been appropriately surveyed. The SDD approach imposes some rigidity which could hinder the outcome, as well as a certain reluctance to refactor or remove anything.

I think the advantages of the software design approach are much more intuitive, although perhaps I will talk about them another day. The takeaway from this project was that it is at least a choice that must be considered, rather than simply rushing into the evolutionary method, as in the end I suspect the two often take an appreciably similar amount of time.