After my previous post introducing Post-Architecture, I received a bunch of positive feedback, as well as enquiries from people wanting to know more. So I figured a follow-up was in order.

To recap: Post-Architecture is a method of defining architecture incrementally, rather than designing it upfront. It prioritizes velocity during the early phases of development and minimizes the risk of making the wrong architectural decisions.

Goals

The main goal of Post-Architecture is to enable a faster, more open and iterative approach to software engineering, including architecture, by avoiding architectural constraints early on in the process.

I think you could say that Post-Architecture is to Architecture what Agile is to Waterfall. Like Agile, it’s not a silver bullet, but an approach that may be beneficial to certain teams in certain circumstances. In the previous post I mentioned startups as an environment where this approach works particularly well. Also like Agile, Post-Architecture embraces the fact that we don’t know everything upfront and we may need to pivot at certain points in time. But where Agile concerns itself with the development process, Post-Architecture defines a set of values for code architecture that works best in combination with such a process.

So the reason I picked Agile for this analogy isn’t coincidental: I think there is a high level of correlation between the two. Are you using Agile methodology or Lean development? You probably want to embrace Post-Architecture as well.

Post-Architecture as a Set of Values

More than anything, Post-Architecture is about a mindset: It’s about being okay with the idea that not all of the code you write is subject to a predefined architecture. With Post-Architecture, architecture trails product development. This is a simple consequence of the values it embraces:

• It’s better to fail fast than to spend more time trying to succeed. When you are early in the process of trying out a new business, or even just a new feature, you may not yet know the final direction you end up taking. The idea for your business or feature may even fail entirely. The sooner you find out what doesn’t work, the less effort you will waste, and the sooner you can try something else that might work instead.

  Even if the thing you are working on does work out, there might be difficulties along the way that need to be resolved before you can succeed. You want to discover those as early in the process as you can.

• Code is a liability, especially during early development. If at any point in the development process you run into difficulties, and you realize you need to change gears, you may need to pivot and adjust your approach. Any code written up to that point may no longer suffice, and you may need to update it, or throw parts of it away. The less code you have, and the smaller your investment thus far, the easier this will be.

• Validate before you commit. Committing to a specific architecture is a decision with a lot of impact, especially when you pick the wrong one. It may feel like a chosen architecture is a safe bet, because it is widely used by other projects. But how do you know it will fit your project? If your requirements are well-defined upfront you can probably still make a reasonably sound bet, but in an agile environment? You better validate your ideas before committing to a specific architecture.

  Implementing a single use case is not enough to validate any given approach. But by the time you’re implementing three similar ones that could benefit from a common abstraction? Now you’re onto something.

• Perfect is the enemy of good. Striving for perfection prevents you from making meaningful progress. Since Post-Architecture prioritizes velocity, it takes the position that progress — any kind of progress — is preferred over time optimizing towards an abstract idea of perfection. It will be hard to find any two people agreeing on the same definition of perfection anyway. It’s okay if it means you need to spend extra time cleaning things up later, as long as it helps you discover more fundamental pitfalls early on.

Taken together, these values should hopefully make it clear why you should postpone major architectural decisions, as well as give an idea as to when it is the right time to start refining your architecture.

Post-Architecture is an Iterative Process

Assuming your product or feature doesn’t fail immediately, you still need to make a conscious effort towards defining your architecture. Post-Architecture is not no architecture.

Someone on programming.dev called it postponed architecture, and they have a point. It very much is postponed architecture, but that’s not the whole story to it. Because by postponing your architectural decisions, it means you will be subject to a different set of constraints compared to if you had defined your architecture upfront. Most notably, potentially a lot of code will have already been written. How can you define an architecture, when the code you’ve written thus far looks like it abides by anything but?

I will give more technical guidance on this in a follow-up post, but the key point is: You iterate. Don’t feel the need to fix everything at once. Make small incremental improvements that fix concrete painpoints. This also reduces the risk of breaking things in the process.

What you want to avoid is that you get to a point where you have to halt feature development for a longer period of time because you are now in a “refine your architecture” phase. It’s hard to speak in absolutes, but usually when an improvement takes more than a few days, I would question my approach and wonder if it can be done in smaller steps instead.

Post-Architecture is Not an Excuse

It’s easy to think that since we want to prioritize velocity, it means we can make a lot of sacrifices to achieve that high velocity. This is not what Post-Architecture is about. While Post-Architecture indeed trades upfront architecture for a postponed architecture, it’s not about sacrificing anything. Post-Architecture is not an excuse.

Code quality should very much be preserved. We should acknowledge that keeping a high level of maintainability across your codebase can actually be harder if your architecture is not yet well-defined. So while Post-Architecture posits that your architecture will become more fixed over time (and assuming your project isn’t canceled, should even lead to a more suitable architecture over time), it means you need other means to preserve code quality while it’s still in flux. Use linters, automated formatters, document your code, draw a diagram if necessary, do proper code review, and especially document anything that feels off. If your code becomes a mess, refining its architecture will be a messy process as well.

Quick ‘n dirty can sometimes be fine, but don’t use Post-Architecture as an excuse to deliver a hackjob. Where to draw the line? As long as someone else can still easily understand what you did and easily improve on it, you’re probably on the safe side.

Post-Architecture is Not a Methodology

I really don’t care if you write Post-Architecture or post-architecture. I also don’t have a defined set of guidelines that you can follow to become a certified Post-Architect.

I think every programmer can operate in the spirit of Post-Architecture if they have the right mindset. To excel in it is merely a matter of experience. And to gain experience you need to make mistakes. I know I made my share. And that’s alright, because this is a mindset that embraces mistakes as long as you learn from them — as long as you keep making forward momentum. Just try to make sure your mistakes aren’t huge ones, so keep iterating in small increments ;)

In a sense, this may feel like a cop-out. By refusing to define a methodology, I effectively refuse to define what good Post-Architecture looks like. If I wanted to bait my audience, I can always refute counter-examples of where post-architecture-like approaches failed and a posteriori claim that those cases wouldn’t fit my definition anyway. A true no true Scotsman defense. Alas, I don’t intend to argue that Post-Architecture is better or worse than upfront Architecture, I merely believe its approach is beneficial in some situations.

You could say that Post-Architecture represents an antithesis to the classical ideas around Software Architecture. There are indeed situations in which all the recognized best practices of Software Architecture can be thrown as caution to the wind. But just as it takes skill to recognize when to best apply a given practice or code pattern, so too does it take skill to recognize when to cast it aside. Post-Architecture is merely here to say you are not crazy for going off the beaten path, to guide you as you do so, and to minimize your risk as you venture out on your own.

Image courtesy of Daniel Paschal

How to Apply Post-Architecture

Thanks for making it this far! If you’re still with me after all this writing, I imagine I’ve struck a nerve and this is a topic you’re interested in. But I can also imagine you’re getting tired of all the abstract talk and are looking for more concrete, practical pointers on how to really apply post-architecture. Some more technical guidance so to say.

I have taken my time explaining the abstract, so that hopefully the technical advice can be placed in the right context. But more technical advice is coming… in the next post in this series. Stay tuned!

Update: The next post in this series is here: Post-Architecture: Premature Abstraction Is the Root of All Evil.