4GLs.

Consider any commodity that can be manufactured. In the early days, goods are handcrafted by specialists who charge handsomely for their time. Good furniture used to be the preserve of the rich who could buy custom built units that would last forever, (with reference to that lovely quote from Alan Clarke about Heseltine, 'the kind of person who bought his own furniture'). Craftsmen were well rewarded for their services and could advance socially because of it. As more people owned property and had disposable income, more people could buy furniture but there weren't enough craftsmen to go around. In any case, it was still too expensive. As there were so many people wanting to buy, entrepreneurs could invest in mechanised ways of creating furniture and mass production was born. The same is true of cars, for example, Henry Ford standardised therefore 'you could have any colour you wanted as long as it was black.'

If you were to plot a curve of skill levels versus cost per hour you would predictably get a standard curve. The very skilled people charge a great deal per hour and there are very few of them. The partially skilled people are relatively inexpensive and exist in the industry in greater quantities. The unskilled have no place in the industry.

It therefore follows that, if the delivery of software is componentised so that the actual skill levels required to deliver a component are relatively low but specialised, both a greater churn and variety of components can be achieved. Take as an example a factory line worker assembling a specific part of a car. The task itself may be complex but it can be learned and once learned repeated ad infinitum. If there is sufficient demand for standardised custom software, not the oxymoron it appears, then it is possible to set up assembly lines of business analysts, DBAs, C++ programmers, VB programmers, ISA experts, IIS experts and so forth. Each level is specialist but can be mastered to MCP level in a matter of weeks--the same cannot be said of truly skilled jobs.

Once at this level tools vendors are able to move vast quantities of software. It is actually in their interests to keep this componentisation as unitised as possible to achieve the highest sales volumes. But having arrived at the point where skills may be acquired rotely, there is bound to be an out-flux of work to where such skills are even cheaper.

Of course, the corollary is that the only people who can afford to develop software using this model are the people who can afford the production line, thus effectively alienating the SME who then becomes dependant on other software factories and must reduce their expectations accordingly.

How can anyone in their right mind believe that a team of five to ten people can deliver a custom solution more economically than a team of one? But that is the myth that organisations, such as Microsoft have successfully marketed and why? Because that means they can sell five to ten pieces of software instead of just one.

So the reason for the resurgence of interest in 4GLs is because it is the only way SMEs can fight back and the only way skilled programmers can see to their job staying in this country. Also, the answer to 'how can a small software house deliver more products faster' lies in another 1980's maxim: "Don't automate--obliterate" The idea is to solve problems by not having them in particular, when you have a lot of layers in a process then the cost of handing off between layers exceeds the benefit of the layers. In business, this led to the 'caseworker' model, where computers brought knowledge to decision makers. In computing this has not happened. 4GL in competent hands could make it so. Microsoft's model of software factory embeds a very high process cost that can be reduced by cheap labour, but cannot be removed.

www.revsoft.com

Andrew McAuley, Revelation Software Limited