Information Age: News, analysis & insight for IT & business leaders

Certainly, when the IT in question is the middleware that integrates the business systems of modern enterprises, anything other than a strategic approach to its management and development is unlikely to be successful.

"Middleware is intrinsic to almost every modern application," says Horn. "Every application is partitioned and needs communications [middleware] to link the front, the middle and the back."

These links between and within applications have to be treated as a strategic issue because once they are in place they often remain so for a decade. They are like railway tracks, he maintains, connecting important commercial centres: "You want to be able to lay them once, you certainly don't want to have to keep digging them up and relaying them."

Unfortunately, today's businesses are dealing with a middleware legacy that reflects the fact that much of the track that has already been built was laid down to solve discrete tactical problems. "There are too many railroad tracks of different widths and gauges," says Horn. He cites an accumulation of proprietary solutions that different vendors have developed over the years, not to mention the vast investment in "dark matter" - solutions that enterprises have devised in-house.

The integration strategy of the future has to be focused on building standard means of linking applications and business processes together, but getting there is not going to be easy. "In some ways," says Horn, "I think the situation today is even worse. We still have all the old railway lines and the dark matter. We also now have three standard [modern development and deployment platforms] - Corba, J2EE and .NET."

Supporting all three of these is not an attractive option for any organisation. But simply choosing one to the exclusion of the others is not an option either, as no organisation exists in isolation from partners, suppliers or customers, who will each make integration infrastructure choices of their own.

The good news, says Horn, is that there are some common threads that will enable these - and many of the existing applications and processes - to work together. Above all, the widespread adoption of web services technology by all the main IT vendors - including, significantly, Microsoft - will prove critical.

While the technology is not a panacea for the shortcomings of systems integration, web services - deployed within the context of a service-oriented architecture - should offer the opportunity for the IT community to develop a common vocabulary for the communication of business and IT functions.

And, potentially, there is much more to come. "The software industry needs to grow up," says Horn, and provide the same degree of predictable and reliable increases in productivity that the hardware business has achieved, as encapsulated by Moore's Law of microprocessor performance growth.

In the software world, says Horn, "this will come about when software starts to be capable of writing itself" - responding automatically to changes defined at the level of business logic. This is still some way off, but web services, says Horn, "are at least a good starting point".