Humans have an infinite capacity for creating disposables, too often to our own detriment. A stark reminder of it was seeing a German milk carton being washed up together with an unfathomable number of nappies on a beach during a visit to India a few years ago.

Why is it that we can produce an infinite amount of these products but software, the one thing that can’t end up in landfill, ain’t one of them?

Software is the sum of ideas and time. It has no physicality (though it does need energy to run, and can waste a lot of that, too). Though it has to run on people’s laptops or servers in a data centre, neither are purpose-built for a specific application and can equally run the next big thing™️.

Whether it’s a COBOL system driving banking services, software used in shipping or aviation – did you ever look at the computer screen as you show your boarding pass at the gate while boarding? – or inventory management software you didn’t even know existed at your local restaurant.

Why have such arcane pieces of software not been thrown away? I see three reasons for this:

1. It works well enough
2. There’s no alternative
3. No one knows how it works

The first is an interesting one. Some software may genuinely do the job without causing major issues. But then it is ‘working well enough’ compared to what?

This takes us to the second reason, lack of alternatives. For those looking for something else this could be another off-the-shelf solution which may not exist, or an in-house development which very few can afford.

I’ll write about the cost of software development another time.

But let’s not forget the last reason mentioned above. Nobody knows how it works. Indeed, software tends to be a black box to its end users as much as it is to those responsible for maintaining IT infrastructure.

Anything written in programming languages that get compiled into a binary, makes it impossible to tell what logic is at play or how certain edge cases are handled. Even if there is access to the source code, open source or otherwise, it leaves us with two more questions:

1. How much does it take to understand what it does?
2. How much of it is really covered by automated (unit and integration) testing?

The question of what it does gets harder as software gets more complex.

And this is where we get back to disposable products. Their nature tends to be single use while software is not.

This is much less a criticism of the software that exists today than a question about the software that doesn’t exist, yet.

They may indeed be a way to make software disposable and cheaply available after all, between off-the-shelf and large enterprise in-house software, beyond what AI assistants like ChatGPT.