Are you coding for change or for stability?

Posted by Marcus Hammarberg on April 18, 2013
Let me tell you a story: when I was in university I took an "advanced" object oriented programming course. This was my first exposure to the topic and I was lost big time. The course was taught in SmallTalk had a very different format; the first day we got an assignment from the professor that ran throughout the 4 week course.

We were very excited since we were going to write a game. An old-school text-input adventure game a la Zork. We teamed up three people in groups and went to the professor smalled crammed room. Here we got the instructions on a single sheet of paper. We almost ran out of there.

Just as we reached the door of the room he called us back (I'm sure he had time that call to perfection):
"Oh yeah, almost forgot. Two weeks from now I will come by and change something fundamental of that game description. Continue as you were."
I've told this story a number of times to system development teams (and some product owners) and the reactions are almost the same every time:

  • There's laughter all around the room. Or giggling at least. Quite often some "Pfffft" can be heard 
  • "Oh man - that's one tall order!"
  • "What a hard professor - that's almost an impossible task"
The thing is: he didn't tell us what he was going to change. Just that something was going to change. In two weeks. 

How do you think we approached that taks? 
We coded very defensively. 
  • "Oh my - what if he decides to change this?" 
  • "This probably needs to be an interface, if we are asked to implement it differently"
  • "No - that needs to be extracted since we then can change that part without touching part X"

Things like that. All in all it was an excellent assignment that thought me a lot about programming OOP and Small Talk. Thank you professor Davidson!

As a consequence we ended up with a very modular system that was easy to change parts of and when that frightful day came when the specification was altered we actually managed to change that in a day and then continue writing cool features like a GUI and scary monsters.

We had optimized for change. Because professor Davidson told us it was coming.

And again

Let me tell you a story: this a story about a system that is developed about 25 years ago. It is a business critical system in a big Swedish company. When I say business critical I mean the products handled in it constitute about 80% of the company revenue.

Right from the bat they where very thorough and kept a great documentation. They also implemented a rigid change process. Changes in this is system should not happen a lot, since that would be quite risky or at least have a great impact if failed.

That scared the company so much that a long array of measures where taken to prevent problems in the system; the code was documented in pseudo-code by someone else than the person that wrote the code. The person that wrote the code was not allowed to test the code (that she wrote).

The same, risk-management measures, where taken when writing specifications. People wrote specifications in several ranks so that things was checked and double-checked before handed over to the IT-side of business.

Fast-forward 25 years and we still have the structure around documentation and risks in place. Things seldom fail but the system is very slow. About 30 weeks from "accepted on backlog" to "delivered in production". There's ca 20 ppl involved in the ranks from idea to specification in the hands of the programmer (that by the way calls themselves constructors since that's all they really do, translate pseudo-code in Swedish to COBOL).

Everything around the setup for this system was to handle risk and changes by aiming for stability. Because we thought that change were going to be a exception. Sadly that's the only thing that the business wants to do: change. And faster and faster. When the system was built there was no talk at all about exposing it's data on the internet. Hey - there wasn't even an internet.

They system was optimized for stability.
Let me also underline that I don't think lesser of the people in the second story. They are great people but they were thrown into a system that was optimized for stability. That's very risky.

Conclusion

These two stories got me thinking.
  • What am I optimizing for when I write my code today?
  • Change will always come. I know that from the outset. "I will change something in that spec during the 25 years the system will live". How do I try to handle that?
  • Do I precieve change as a risk or an enabler? Being able to change fast is a business advantage and a way to manage risk. How do I make my code being easy to change?
  • What kind of documentation do I need in a system that I'm going to change a lot?
What do you optimize for?

I wrote a follow-up post on this topic if you want to read more.



Published by Marcus Hammarberg on Last updated