The Many Facets Of Quality
Some time ago, in a talk, someone asked "What is software quality, what gives quality to software?". My immediate half-joking answer was "Software shouldn't get in my way, that's quality!".
It was a half-truth, or a quarter-truth, because really, software quality is something with far too many sides.
We all want quality, but how can we tell what quality is?
Quality is somewhat of a magic word, it means everything yet precisely because of that it means almost nothing.
We could define quality as "what is good" or maybe say that a particular piece of software has quality if it "tends towards what is good and avoids what is bad". But then we've shifted the mystery towards the already fuzzy concepts of good and bad.
In my previous article, I outlined that software must be understood in terms of its "HOWs", "WHATs" and "WHYs", which loosely map to:
- HOW the computer system works below the hood: The inner details.
- WHAT is the interaction between the computer system and the human: The dialogue.
- WHY is the human driven towards the system: Human will, which shapes our systems and our relationships with them
And quality being what is good, it is what's good for every human in relation in the system. Quality permeates every layer but ultimately springs from the WHY.
A few examples:
- We developers want the code to be harmonic, beautiful, well tested and properly structured. We call that quality. But why? Because this quality in the code means it's easier to reason about the system, it will be less likely to have bugs, which annoy the end user and cost time and money, it will be easier to add new features, it will be easier to add new developers to the project, etc. (And of course there is a certain drive towards beauty that can possess developers, when the code quality becomes not a means but an end in itself, like the enjoyment of a fine wine or a symphony... even in that case, it rests on a WHY, a self-sufficient one)
- We want the user interface to look pretty, to be clean, not to be cluttered with superfluous information and to have important information hidden from sight, etc. But why? It might be for purely aesthetic purposes, or you might want to improve the flow of the user and increase productivity or to avoid mistakes, or to improve adoption in your app, to free time for your user, etc.
(Note here that I've made no deep down distinction between quality and value. Some people use quality and value indistinctly, or use quality to mean structural, internal quality, i.e. quality in the HOWs)
So, to understand what quality is made of, how to create it, and appraise it, we're back to understanding the HOWs, WHATs, WHYs and ... that is, the computer system, the human system, and their intersection.
The Right Thing vs Worse is Better
Richard Gabriel in a famous (at least in some circles) essay titled The Rise of the Worse Is Better posed two contrary 'styles of design' characterized by the phrases 'The Right Thing' and 'Worse is Better'. Letting aside the actual contents of the essay, I think the phrases themselves capture a powerful idea, and I want to convince you that worse is, indeed, better.
Why? Because quality:
- is complex and multifaceted
- does not exist in a vacuum, it has its 'enemies': constraints.
That quality is complex and multifaceted shouldn't be a surprise, but it is easy to forget with the mindset of 'The Right Thing'. Quality is clear, expressive code, but it's also being secure, having good traces of execution, being scalable, having low latency, solving the needs of business, etc.
Quality is also to discover what is really needed: A process of "good quality" should be (theoretically) better at figuring out business requirements that needed to be tackled that in a lower quality process might be forgotten or neglected, while correctly figuring out which features and corner cases can be safely de-emphasized.
Quality also has constraints. 'The Right Thing' mindset can lead some people into thinking every feature is worth implementing, every refactor is worth making, every bug is worth fixing. But work requires time, money, and focus, and competes against other possible actions for resources.
(Note here that I'm including everything inside the concept of quality, not only the end result, the quality of the running code and the visible system to the end user, but also the process associated with building and using the software, and the costs associated with this process)
When, instead of building software, we select software, we risk falling prey to similar partial views of quality. For example, a language might have fancier features than another language, yet be beaten in terms of performance, or available libraries, or IDE integrations, etc.
We can now understand what 'Worse is Better' really means (if you read Richard Gabriel original article, that's not his interpretation, but it is mine). It means that 'what seems worse from one perspective might be better from another, while The Right Thing might be... wrong'.
Now, certainly, one can do bad tradeoffs "in the name of tradeoffs". Maybe worse was really worse after all. The key lesson here is to not get caught in "obviously correct" points of view. The "obviously correct" point of view can seduce us with a consistent and clear explanation of why X is better than Y, but could also be ignoring a lot of other facets that could also be important to take into account.
So, I haven't really told you what quality is, right? I couldn't if I wanted to... at least not in a short blog post. Quality is a fractal, many facets that when looking closer reveal more aspects to be taken care of.
So, for example, we want code to be maintainable. When we look into what makes code 'maintainable', we find more 'facets of quality': has good expressive names, logic is encapsulated and isn't duplicated, state mutations are easy to track, objects and modules have uncluttered interfaces, code is easy to navigate with the aide of an IDE, entities are reified, we have good overview documentation of the system, non obvious segments of the code have in-place comments, logging allows you to trace activity and find the root cause of bugs, etc. And each one of these aspects, in turn, matter more or less to different people, accustomed to work with different flows, different tools, different mindsets. And in code bases where the code isn't perfect and people are often running against the clock and the budget, tradeoffs in quality are often made (good tradeoffs? bad tradeoffs?). For some people, "quality" will be finishing the project in budget.
In summary, we can achieve quality when we admit is not an easy thing to achieve and we probably don't know what it is or how to achieve it. To fully know what quality is and what makes it, we would have to be omniscient and know every little detail of every interaction (and what drives those interactions) each human (user, developer, ops, customer, etc) has or can have with the software. Quality must remain 'open' for us in order to truly get it.
What does 'open' mean? Nothing more than this: Acknowledging that our understanding of what quality is, is incomplete. Only by accepting this, then we can have an honest conversation about what is good and bad for each one of the stakeholders, what are the constraints and the tradeoffs we're willing to make, and make plans for achieving it.