Follow Simon Caulkin on Twitter



Article Archive
2021
August
July
June
May
April
February
2020
December
October
September
August
July
June
May
March
February
2019
December
November
October
September
August
July
June
May
April
March
February
January
2018
November
October
September
August
July
June
May
April
March
February
January
2017
December
November
October
September
August
July
June
May
April
March
February
January
2016
December
November
October
September
August
July
June
May
April
March
February
January
2015
December
November
October
September
August
July
June
May
April
March
February
January
2014
December
November
October
September
August
July
June
May
April
March
February
January
2013
December
November
October
September
August
July
June
May
April
March
February
January
2012
December
November
October
September
August
July
June
May
April
March
February
January
2011
December
November
October
September
August
July
June
May
April
March
February
January
2010
December
October
September
2009
November
October
June
May
April
March
February
January
2008
December
November
October
September
August
July
June
May
April
March
February
January
2007
December
November
October
September
August
July
June
May
April
March
February
January
2006
December
November
October
September
August
July
June
May
April
March
February
January
2005
December
November
October
September
August
July
June
May
April
March
February
January
2004
December
November
October
September
August
July
June
May
April
March
February
January
2003
December
November
1998
January
1997
January
1996
February
1994
May


Software must stop bugging us

Sun, 7th Mar 2004

PICTURE an industry that grew fat in a postwar sellers' market. Its products were flashy, poorly built and accident-prone but that didn't matter, at least to producers, because they had obsolescence built in and everyone needed them.

Everyone knows what happened next. The quality message preached by Western gurus, ignored by home manufacturers, was eagerly adopted elsewhere. Soon, foreign firms were invading the market. Jobs started to haemorrhage overseas. While Western manufacturers squealed about cheap imitations and low labour costs, in fact they were losing out not to cheap but to smart. It wasn't an economic but a management story.

The motor industry, of course, is a well known study in corporate complacency. Only we're not talking about cars in the 1960s. We're talking about software today.

The parallels are as close as transistors on a microchip. As computer users know full well, today it is commercial software that is unsafe at any speed - just like the cars of 40 years ago. For Detroit's Big Three car-makers read the software giants that today rule the computer world, such as Microsoft and Oracle. For Japan read India. And for W. Edwards Deming and J.M. Juran, the quality sages whose message Western car manufacturers scorned until the eleventh hour, read Watts Humphrey.

Humphrey, 77, spent 27 years at IBM, where he was director of programming and ran the software labs. Since 1986, he has been crusading for better software at Carnegie Mellon University's software engineering institute (SEI), where he is a fellow.

Like Deming, to whom he is often compared, Humphrey insists that it's perfectly possible to build defect-free software - but not on today's craft methods, which (crudely) consist of relying on individual effort and intuition, and testing out the inevitable errors afterwards.

Wrong, says Humphrey. Writing good software requires method to ensure that quality is built in from the start. Far from being more expensive, he says, building defect-free software saves both time and money because it does away with rework. One estimate is that 75 per cent of Microsoft's software costs are in testing and correcting defects. IBM had two acres of computers just dedicated to testing - all generating cost, not value.

Again like Deming, Humphrey refuses to blame programmers for the errors that cost so dear. To blame, he says, are the management systems within which people work. To prove it, he has developed personal and team-based software development methods that yield results that are remarkably different from the industry norm.

Whereas half of traditional software projects take more than twice their allotted time or are simply cancelled, with vast cost overruns, a recent SEI study of 20 projects run on Humphrey's methods found that they all came in on time and on budget, essentially defect-free.

None of this is new. In 2000, Cheryll Barron, a past Silicon Valley correspondent for the Economist and Business Week, wrote a much quoted essay on salon.com called 'High-tech's missionaries of sloppiness'. Airing just these concerns, Barron railed at computer failure rates running at 25 per cent a year and prevailing attitudes summed up in notably explicit fashion by industry evangelist Guy Kawasaki (also quoted in these columns) as 'Don't worry, be crappy'. The important thing, Kawasaki said, was to ship product fast. If it was 'good enough', problems could be fixed later.

The trouble is that 'good enough' is not good enough any more. Although Barron's piece caused a stir at the time, nothing much in the industry has changed since - except in two vital respects.

First, the stakes are higher. Windows now contains nearly 100 million lines of code. By SEI's rule of thumb that means tens of thousands of errors even after testing, which by definition won't be discovered until an accident happens, or worse.

In the firing line now are more than individual users. Unlike cars, computers are interconnected by the internet. As size, complexity and numbers of interconnections grow, so do the vulnerabilities of the system as a whole - which can simply no longer be addressed by the industry's traditional reactive methods. Baldly, internet security problems are software quality problems, says Humphrey: 'It is safe to say that there is no such thing as poor-quality secure software.'

The second change since 2000 is in the competition. Echoing Japanese car firms of the 1960s, the most eager adopters of software quality systems are not (with exceptions such as IBM and Boeing) US industry leaders but much less prominent Indian firms which have an urgent need for an edge - and no vested interested in the past. Indian companies such as Infosys and AIS are now churning out code that is not only cheaper but better than the industry standard. Of the organisations claiming the SEI's (top) Level 5 quality status in 2002, more than half were Indian. Most US and European firms are at Level 2.

Back in the 1960s, the most assiduous motor-industry follower of Deming's principles was Toyota, which has never stopped deepening his insights and honing its production system. Forty years on, guess which is by far the most profitable car firm and about to overtake Ford as the second largest in the world.

Could it happen in software? It sounds improbable now - but then, that's what General Motors thought in the 1960s. Most firms may not know it, but the process has already begun.

The Observer, 7 March 2004


<< Back to Index

User comments

Name:
Comment:
Check:4+1=