Software Can’t Live On Its Own

We’re building software in hope that some day we’ll leave it and it will live on its own. Or with minor supervision. But the other day when my father asked me to dig an old website, I did some thinking and realized auto-pilot software is almost never the case.

Software is either being supported, or is abandonware, or is too simple. We constantly have to “fix” something, on each piece of software. Basically, picking up an old project and running it is rather hard – it would most probably require upgrading a ton of components. For example:

  • Fixing edge cases, bugs, security issues. The software environment is dynamic, and no complex software is without uncovered edge cases. Security issues arise constantly and have to be patched. Unless we find a way to write bugless software with perfect security, we have to support all these.
  • Breaking upgrades:
    • browsers are being upgraded constantly, and old websites probably won’t work. Protocols remain backward compatible for a while, but then support is discontinued and one has to upgrade. Operating systems introduce breaking changes to software running on them – one clear example is Android, where with each major version something doesn’t work anymore (because it was deprecated 2 versions prior) or has to be done in a different way. We have to be there and tweak our code to accommodate these upgrades.
    • Frameworks and languages get upgrades as well – and sometimes we can’t even build our legacy software anymore. Even if we can, the target environment may not support our old versions. The aforementioned site was written in PHP 4. Shared hosting providers no longer offer PHP 4, so will that site work? Possibly will need tweaks.
    • Changes in 3rd party APIs. If you rely on something like a facebook API, or a Google API, chances are your 3-year-old project will no longer work.
  • New use-cases – the real world is dynamic, and software that supports some real-world activity has to change with it. Some features become obsolete, new features are needed. Vendors like to advertise “draw-it-yourself” tools that create new forms and business processes without any technical expertise, but that’s rarely working properly
  • Visual design becomes outdated. Remember Web King? Maybe that was the design of the 1995, but not anymore. We’ve gone through waves and waves of new design trends, and often it’s not okay to look outdated.

A piece of software is not like a building – you can’t build it once, and leave it exist for decades, with just occasional repairs. It is not like a piece of kitchen appliance, you can’t just replace it with a newer version.

It isn’t like a building, because it’s too complex (not diminishing the role of real architects, but they have a limited set of use cases). And it isn’t like kitchen appliances, because kitchen appliances don’t have data.

And actually, data migration is one of the reasons legacy software exists – migrating it to something new is hard – one should fit it into a new structure, and into a new database. And even simple migration from an older database version to a newer database version is hard. Migrating structures and even usecases is horrible. I won’t even mention triggers, stored procedures to be migrated across vendors and so on.

So yes, keeping an old piece of software running requires a lot of effort; migrating it to a newer and better piece of software is often a doomed project and you’re stuck with your existing system forever.

That means there is a whole big branch of the IT market that focuses on that – providing software to clients and then keeping them bound to that software forever. With regular updates and support. There is another type of companies, where things are more straightforward – the “single product as a service” companies. The cool web 2.0 startups are mostly single-product-as-a-service companies and if the company dies, the product dies with it. If the company manages to make some money, you don’t care…until it dies, and then your migration to a new piece of software is the promised nightmare.

Leaving simple software aside (my computoser is running unsupervised for 2 years already; not that it’s simple, but its complexity is confined to the algorithm; I heard that the software for the trash cleaning company I’ve written when I was 16 is still in use in my hometown) everything else needs constant caring. And given that more and more software is being build, this leads us to the sad realization that we’ll have to support a lot of software. More and more of programmer’s work will be caring for what’s been already built, rather than building something new. And on one hand that’s sad. That means software for many is not “craftsmanship”, not “science”, not “making cool things”. It is a mundane support and gradual extension of old, clunky bulks of code.

Unless we learn to build self-supporting software. Software that automatically overcomes OS upgrades, framework and protocol upgrades. Software that allows extending without writing code (which many systems claim to do even now, but very few actually do). Until that time, I’m afraid we’re stuck with supporting our current projects, and in the best case – extending them to fit new needs and customers.

2 thoughts on “Software Can’t Live On Its Own”

  1. that’s why i try to keep as much stuff as i can in configurations. my code is a black box with a bunch of algorithms which crunch the config to generate an output. code change does not really affect me much.

Leave a Reply

Your email address will not be published.