It was Ray Davies of the Kinks who wrote the satirical pop ballad "Dedicated follower of fashion". He was inspired by the London youth of the 1960s slavishly following Carnaby Street fashion trends. However, he could as easily have been observing the IT industry. For me this was best exemplified when I worked as a marketing manager. In the space of four years I went from being the Unix marketing manager to the Open Systems marketing manager to the client/server marketing manager. Throughout these title transformations, my duties were constant: to help my company market Unix boxes.

Five years ago client/server was the great future of IT visionaries. They said we could leverage the existing desktop investments by enabling PCs to execute part of the business applications. This would empower business users by giving them closer proximity to their operational data while, at the same time, reducing the significant investments being made in running data centres.

The reality was a bit different. Everyone overlooked the high costs in managing a dispersed environment. Furthermore, network performance sabotaged the promise of deploying distributed databases. However, while the euphoria of client/server has subsided, the industry has been moving to address those concerns. In fact usage of client/server architectures has increased significantly over the last five years. In IDC's Forecast for Management survey in 1994 just under 40 per cent of respondents indicated they had implemented client/server. In the 1998 survey 70 per cent of organisations responding had done so.

What is interesting is that the implementation styles of client/server have changed noticeably. In the 1994 survey 17 per cent of respondents were deploying client/server to modernise legacy applications through a Windows and GUI screen interface. This year the figure dropped to 8 per cent. Conversely, over the same survey periods, the percentage of organisations with all the data residing on the servers and all the processing logic based on the clients has increased from 10 per cent of respondents to 21 per cent. This would seem to imply that some of the network challenges of client/server have been addressed to the satisfaction of an increasing number of CIOs.

This is supported by research conducted in the US last year by Computer Economics, which surveyed CIOs to determine IS architecture trends over the last three years. The results indicated that totally centralised IS architectures had declined from 12 per cent of organisations to a mere 7 per cent. On the other hand, the percentage of organisations deploying a mixture of centralised and decentralised architectures had increased from 27 per cent to 39 per cent of respondents. Nevertheless, in the same survey the number of organisations utilising a totally decentralised IS architecture had remained static at 12 per cent. Clearly not all client/server reservations have been addressed.

But the most recent newsletter from Computer Economics had some encouraging news for client/server. It examined the cost of ownership of client/server and concluded that its outlook was definitely bright. The newsletter observed that there had been a steady climb in budgets allocated by CIOs to client/server. It believed this was happening because there was a growing probability for a good return on this investment. The likelihood of a favourable return on a client/server investment had risen from 56 per cent in 1996 to more than 70 per cent today. Computer Economics attributed this to the fact that CIOs were leveraging the knowledge gained from earlier, more costly, deployments of client/server.

While the cynic in me might poke fun at the fashion followers in IT, perhaps there is another side of the coin. Our knowledge and understanding of where technology can be better implemented is constantly evolving. Concepts that were once impractical can be found to still have application when new functionality is available to complement them. As Thomas Edison once said: "I didn't fail in 1000 experiments before I invented the light bulb; I just discovered a 1000 ways it wouldn't work." Peter Hind is the manager of User Programs, which includes InTEP, at IDC Australia