Tuesday, February 28, 2012

Technical debt - developer moans

A few days ago I pushed out a couple of Tweets which got quite a reaction:

“I am fed up hearing developers complain of Technical Debt as if it has nothing to do with them. Who writes this stuff?”


“Devs say 'look technical debt' like 'look its snowing'. As if it has nothing to do with them”

It was that second one that got a hail of re-tweets.

So I thought I should add a few words of explanation here.

First it is important to say that the these tweets were not aimed at, or about, anyone in particular. I’ve visited a bunch of companies and delivered a lot of training in the last few months and the topic of technical debt comes up a lot. If I had anyone in mind about these comments in particular it was a developer I worked with four or five years ago but his sentiments are common.

Second, I’m not sounding off about developers. I love developers, I used to be one. I’ve had the same feelings myself at times, speak to people who managed me way back when and I probably moaned about the same things. (Over 10 years ago I wrote High Church C++ where I agonised about the right way to write code.)

What I am taking aim at is an attitude which says: “There is this thing called Technical Debt, we have it, it makes our lives hard, we want it to go away, but we have nothing to do with creating the problem and we need a special solution.”

It sometimes seems to me that having labeled the phenomenon developers are too keen to use the label. Steve Freeman and Chris Matts have tried to recast the debate in terms of real options “Bad code isn’t Technical Debt, it’s an unhedged Call Option”. Intellectually I agree with them but I find that the explanation required to explain an “Unhedged Call Option” is greater than the explanation to explain the original problem so I’m not sure what the metaphor adds to the debate.

The other problem with switching to an “unhedged call option” metaphor is that it perpetuates the Quick and Dirty Myth - the idea that doing a bodge-job now will pay back in shorter delivery time. In reality the reverse is just the case, I gave some statistics on this in my blogs last year about Capers Jones book Applied Software Measurement.

Common reasons given for the existence of technical debt are:
  • It was the developers who where here before us
  • We don’t have the time to fix it
  • The managers made us do it, or The business demands this
  • Its the Architects
(Notice the current developers are never to blame - and as much as I don’t want blame thrown around thats the debate is too often framed.)

Anyway, none of these reasons really stand up to analysis, except perhaps the last one.

Regardless of what the developers who were here before did the current developers should be addressing the issues. It is the current developers who need to live in this code, if they don’t like it they should do something about it.

Sure there may not be time to stop the world and fix everything but you don’t need to. You just need to fix a little bit every time. As Uncle Bob Martin says in Clean Code, follow the Boy Scouts Rule “Leave the code slightly better than you found it.” Then let the Power Law work its magic.

Secondly, there is more time if you do it right. Building quality, bug free, systems is faster than the alternative. Forget the Quick and Dirty Myth, its Slow and Dirty or Quick and Right.

(Sometimes I meet developers who want to “stop the world” and re-write the system. I’ve also on occasion met managers who have bought into this argument and stopped the world. This is a seriously bad idea, its why Netscape no longer exists: developers were allowed to re-write Navigator, as they were doing so Microsoft attacked.

I keep meaning to blog about “version 2” now I must, just watch this space.

Twice I’ve come across Managers who told the team they could re-write. In the first case the managers were lying. The developers, well one specifically, held a gun to their head and said “I’m resigning on 14 February if you don’t let me rewrite”. In that situation you should let them resign, these managers were scared of the developer. As far as I can tell things didn’t work out, the developer left the company not long after and the company was sold.

In the second case the Managers were either stupid or desperate. In either case within days of starting the “Technical Improvement Programme” (TIP) they found it wasn’t that easy and things started to revert to usual.)

Getting back on topic...

Unless Project Managers and Business Analysts sneak into the building at night and change code when developers aren’t looking then managers do not write code. Nor does the nebulous unit called “the business”. It is developers hands on the keyboard, “I was only following orders” is not an acceptable defence.

There is a professionalism issue here. Software Engineers need to take their work seriously and create code they are proud of. This is not to say they should start gold plating things but being an professional engineer means you don’t do shoddy work.

Architects may be the only defence I might accept: however this defence is only applicable when the organisation and architects are committing an even bigger sin, namely Architects who Don’t code, or Architects who don’t understand the system - usually these two go hand in hand.

If the architect(s) code then they are part of the development team, they see the same problem, they have the same complaints. Now it is just possible that the architect does code and they aren’t very good. Or rather, the Architects understanding is different to that of the developers, which might point to issues of recruitment, the architects style or bad team dynamics.

The reason I might accept this reason is that it points to an even bigger problem for the organisation and it does imply engineers are pull in opposite directions.

Interestingly Ron Jeffries replied on Twitter, and said “Ward [Cunningham] originally defined tech debt as reflecting later learning, not available at the time of construction.”

I think its time we returned to this definition. Its not about other developer, the business, architects or others creating technical debt for poor developers. Its about developers improving their own understanding. Its about the need for refactoring. And it most definitely is the developers own responsibility and professionalism.

1 comment:

  1. First - I think you meant, "alternative" but whatever you typed got corrected to, "alliterative" - which makes for amusing reading :-)

    I agree with your overall point - which is that "technical debt" is frequently used to refer to something not in the spirit of the original term.
    May I propose, "technical deficit" for this instead?

    I don't completely agree with all the details, however. For example, while not true in all cases, sometimes developers really are given strict instructions to use practices that result in low quality work - where the alternative is to be out of employment. As long as real lives are not at stake the war crimes metaphor doesn't really hold up.

    I have been in that situation a handful of times. Often I have ended up leaving. In at least one case it wasn't feasable for me - at least straight away - so I tried to do what I could "from the inside".

    - Which brings me to the second point. Applying "the boy scout rule" when the reasons for code rot are systemic is (or can be) a losing battle. Bad code is being written against bad designs faster than you can clean anything up.

    Of course it comes down to whether (having tried all courses available to improve things, of course) you can stay in that situation. That depends on a lot of things - professional pride is only one of them.

    I'll touch on how to reconcile all that in my ACCU keynote, btw.


Note: only a member of this blog may post a comment.