November 21, 2012

Who's Afraid Of Legacy Code?

British society has a bit of a problem with age. We're culturally obsessed with youth, and prefer to hide away our ageing population - presumably so they don't keep reminding us that we, too, will one day get old.

This paradox leads to severe consequences for society, as we choose to ignore the facts of life and live as if old age will never come. We don't look after ourselves as well as we should, we don't plan for our futures as well as we could, and we structure our society to favour youth in many respects. That our society is gradually getting older on average, with over 65's now making up a big chunk of the population, confounds our desire to live in a youthful world. Ironically, this growing generation of senior citizens has led to successive governments pandering to the older vote at the expense of the young. Young people are now paying net for an older generation who have not adequately provided for themselves (chiefly because nobody thought people would be living as long as they are), and the "grey vote" has become such a powerful block that young adults of tomorrow can expect to be shouldering even more of that burden.

Software development has grown a similar paradox. We're obsessed with the shiny and the new, despite the fact that we're surrounding by a growing legacy of old code.

Nobody thought that the software they were writing back in the 90's, or the 80's, or 70's or 60's, or even the 50's, would still be in use today. And so, they didn't plan for the future we now find ourselves in.

When you open up a book or a magazine or read a blog post about software development, chances are it will be about writing new code. Aside from some noteworthy exceptions like Michael Feather's Working Effectively With Legacy Code, most coverage of software development is about programming on a blank sheet.

This has a two-fold effect; firstly, most developers lack the skills and the disciplines needed to maintain or add value to existing softare. And secondly, most software is not written with a potentially long life in mind.

Not only do developers lack the skills for legacy code, they have a marked tendency to run a mile in the opposite direction from acquiring those skills. I run a training company, so I know how low the demand is for learning them.

Employers, too, fail to recognise the need for and the value of legacy code skills. They rarely ask for them when hiring developers, and tend not to support developers seeking to improve things in those areas. When did you last see a job advertisement asking for experience of restoring or rehabilitating old, knackered code?

This is despite the fact that most developers are working on legacy code, and that their inability to add value to it and respond to the changing needs of the business and the end users is often cited as a major barrier to business competitiveness by managers.

As Ivan Moore recently put it, legacy code is "the elephant in the room". It comprises the bulk of the work and the overall cost of software development, but occupies a minimal slice of our thoughts and our care.

In the last decade, Test-driven Development has become de rigeur. And a jolly good thing, too.

But, as I see time after time, it's entirely possible to produce legacy code doing TDD. And even if you master writing clean, maintainable code, what about all the code you already write that's out there serving users right now? Do we just write those huge investments off to experience?

No. That would be silly, and immensely wasteful.

If only for the learning experience of dealing with the consequences of design decisions - I've yet to meet a genuinely great developer who hasn't devoted significant time to cleaning up old code - it's high time for us to really get to grips with legacy code.








Posted 2 years, 3 months ago on November 21, 2012