Premature Architecture

A few months ago, I started learning Flutter to build an app I’d been thinking about for a while. I joined the various Flutter-centric communities and followed developers who were active in the space. Overall, I’ve thoroughly enjoyed both the framework and the community dialog. But one element jumped out right away: a tremendous amount of material put out in social media and spammy Medium posts bikeshedding over what I considered “big software” topics like architecture. I was taken aback since most of the Flutter usage was for building a mobile app, and it was often done by a single developer. Whatever gets these people over the finish line is what they needed, so all this Architecture discussion seemed like quite a distraction. There has been some pushback, for example, this excellent tweet thread:

https://x.com/peteralexbizjak/status/1783015772463263938

(Apologies to those without a Twitter account who can’t view the thread. Thanks Elon.)

This topic has been gnawing at me enough to actually write a blog post, but first, a little history about how I was introduced to programming.

Foundations Link to heading

We got our first computer in 1986, and it looked like this:

(Credit)

Sitting at a DOS prompt and then starting GW-BASIC was pretty much how everything began for me for quite a while. I just wanted to make the computer do things (like, for example, drawing a line on the screen), and the resources I had were the inscrutable system manuals (the binders to the left), one book on BASIC from the now-defunct Borders bookstore, and a lot of time as a kid. Importantly, I didn’t have the entire internet telling me the right way, best practices, don’t do that, OMG did you really just, etc. etc. I just wrote code however the hell I wanted until the program worked like I thought it should. The code was just a means to an end, which was a working program.

As I progressed, got a modem and access to more material, and got some more books, I started learning some valuable CS ideas that undoubtedly helped. These were important because they usually made the program run better (e.g., don’t read from floppy so much, cache some data!). Still, though, the rules were few. I used goto and global variables with gusto, wrote 1000-line functions, and definitely didn’t even know the word “Architecture” could apply to software. But the programs got written, they worked, and I had a ton of fun. Fun is important, by the way! When we give a kid a bucket of Legos, we don’t also say, “Here are the best practices for assembling these. Please build Legos this way.” We just encourage them to build.

Premature Architecture Link to heading

Back to Flutter. When I started seeing big discussions about architecture (Service Oriented Architecture, Domain Driven Design, Clean, etc.), I was surprised because those are things I’d associated with large systems being maintained by many people. Uncle Bob agrees:

The goal of software architecture is to minimize the human resources required to build and maintain the required system.

When a group of people, especially a large group with people coming and going, needs to collaborate on a project, the coordination and integration of their work becomes a dominant factor, and architectures can help provide useful guardrails. But what about the solo developer (or tiny team) building a mobile app?

I remember learning about the Small-angle approximation and though that it was somewhat magical. In short, for small angles you can often approximate certain trig functions, e.g. sin θ ≈ θ. You can find many similar “for small n…” type rules-of-thumb in the CS world, where for small cases there may be a simpler option, and I think most of us (myself included!) don’t heed them enough. If we’re talking about a solo developer building a mobile app, we can comfortably start with that being a “small n” situation, and choose accordingly.

Back to Uncle Bob. Since we’re talking about “minimal human resources” (developers) working on the app, unless any architecture-focused work sufficiently helps that person build the app, it is probably not worth the effort. Similarly, the maintenance of a mobile app by one person is likely not going to be significantly impacted by an architecture choice since most apps–if they ever ship–will only see minor bug fixes or straightforward feature additions. Folks aren’t going to replace a working database in their working app just for fun, and having an elaborate data store abstraction probably won’t justify the development cost. Just get the thing built, and remember, all those architecture decisions are compiled away, and the end user only cares if your app works.

It’s all tradeoffs Link to heading

This cavalier approach is all good, but what about the real world, with companies, interviews, existing code bases, and all that? Of course, you have to adjust to the situation presented. If you want to get a job in places that demand fluency in design patterns and enterprise architecture, then you have to be able to deliver on that, both to get the job and because it probably makes sense in that environment. You should follow an existing code base’s design, which may mean learning how these architectures work.

Likewise, if you enjoy the elements of architecture and derive joy from your code being structured just so, then by all means, go for it! I, for example, really got into OO programming and thoroughly enjoyed trying to leverage patterns from the “Gang of Four” book. The problems arise when you’re not producing because you’re spending your time reading Medium articles and are paralyzed trying to pick a state management solution.

If you’re getting started and want to build an app, get it built however you can. Your #1 enemy is loss of momentum, so make progress first and foremost. (No extra points will be awarded if it looks like Enterprise Fizz Buzz.)

And remember, right now, someone is getting paid to ship spaghetti code with jQuery v1, no tests, and it’ll probably work just fine.