David Norton

A member of the Gartner Blog Network

David Norton
Research Director
7 years at Gartner
25 years IT industry

In his role as research director with Gartner's application development and architecture team, David Norton supports clients by developing and delivering quality and timely research. Read Full Bio

Coverage Areas:

Want to see the perfect component? – just open your eyes.

by David Norton  |  September 25, 2009  |  1 Comment

Monday night saw me settle down in front of the TV to watch “What Darwin Didn’t Know” on BBC. A documentary described as “the story of evolution theory since Darwin postulated it in 1859 in ‘On the Origin of Species’.” Towards the end of the program, and just before I nodded off (having two boy’s under 4 staying up past 22:00 is quite an achievement) the presenter introduced the tree of life and the DNA evidence for it. OK nothing new there, but then he brought up the subject of shared genes. Yes he said the usual stuff on humans and chimpanzees sharing 98% of their DNA. But he also went on to give a very specific example – a gene pair called Pax-6.

Pax-6 is a control gene; it triggers eye development in human embryos. It also triggers eye development in chimpanzees, apes, mice, rats, cats, bats, and fruit fly’s. The Pax-6 gene triggers eye development in basically every creature that has eyes. So even though the eyes in a spider and a dog look very different they have the same starting point.

But there is something more important about Pax-6 – its old, very, very old. Over 500 million years ago sometime in the Cambrian period the first proto-eye developed. Those early eyes just detected the presence or absence of light, a simple but major evolutionary milestone. And Pax-6 was there, the trigger for that first proto-eye.

Pax-6 is so good at what it does that is has been passed down and across species unchanged. It’s position within species chromosomes changes but its still Pax-6. Its replication across the species is so good that you can take the Pax-6 of a mouse, transplant it into a fruit fly embryo which will then go on to develop eyes as normal. Genes for arms and fins have come and gone (and in some cases come back again) but Pax-6 remains.

So at this point you probably think thanks for the lesson on gene evolution but what has this to do with a software components? The answer – Pax-6 is an example of a fundamental building block in nature, a tiny component. It’s simple but so perfect that its on every creatures top 10 list of genes I like to have – ultimate reuse.

Could you imagine the equivalent of cross species reuse in IT? Taking a core component of SAP or Oracle and dropping it into every package application, regardless of version, on the planet and expecting the component to still work!!! We still have trouble doing that within the same package, even when we design for it.

The most impressive thing about Pax-6 is it was not designed to be this uber-gene with the power to cross species. The first creature to have the gene did not evolve it as some sort of altruistic gesture designed to be reusable by other animals. The reason why its still doing the job today is because its simply the best way of getting the job done.

When developing services and components we talk about designing for reuse. Maybe we should take a lesson from Pax-6, and focus less on designing for reuse and focus more on designing the simplest, most stable component for the task. And let reuse come from the component being selected and pulled because it’s the obvious choice, not pushed (rammed down your throat in some cases) into the solution.

Now I can hear you say “But Dave one of the principles of evolution is mutation may do nothing or even harm the species – and nature may have many attempts at getting it right” Yes that’s true our tiny Pax-6 component was probably not the first attempt – the control gene that placed a eye on your tail was never going to catch on. But that highlights another problem with IT. You either try and design the perfect “gene component” from the start which is a bit like trying to create a complex multicellular organism from nothing. Or you start with basic building blocks and evolve, adapt, add, and take away over time. The former has risk of outright failure the later has risk of individual “component mutations” failing but higher probability of overall success. The above is a long winded way of saying incremental change is better than big bang.

We could take this analogy further saying gene evolution is an example of an open system, with feedback – adaptive to changes in the environment. Individual sub-systems and components may be unaware of their outer environment but the system they are in is. In IT we often ignore that feedback or react to it too late, we need to make a conscious effort to seek out and act on feedback. This is also a guiding principle of agile and adaptive systems development.

So what can a billion years of evolution tell us? Focus on doing the job in hand, reuse needs low level simplicity even when dealing with complex systems, and complex systems do not just come into being, they are built up over time based on continuous feedback.

1 Comment »

Category: Agile Application Development SDLC System of Systems     Tags:

1 response so far ↓

  • 1 Nick Jones   September 25, 2009 at 6:55 am

    Or maybe the message is: use more genetic algorithms (GAs) and genetic programming (GP). At least then we’d be using a process that has a track record of generating useful components. Unlike some methodologies out there. (just joking… I think).