A new survey from CA (the former Computer Associates) with Vanson Bourne -- The Mainframe: Surviving and Thriving in a Turbulent World -- suggests 83 per cent of UK IT chiefs expect to have problems with mainframe skills while 43 per cent are spending on training to address the issue.

I'm not sure when it was exactly because these things are never exact, but around about five years ago people stopped writing obituaries for the mainframe. These 'Dear John' letters had been around since 1990 at least: 'the mainframe is a dinosaur' they said; 'the mainframe era is dead, long live client/server'.

But the mainframe is still whirring away in the back office and sales are not even declining. It has become the ultimate e-commerce back-end at a time when availability and reliability are more prized than ever. It is the ne plus ultra in security at a time when the threats are more menacing than ever. But still, skills are declining.

Why is this? Well, the answers are necessarily multiple. First, we know that IT on a whole is declining in popularity as a degree subject. Second, the mainframe is not seen as attractive or modern. It was at its peak when the Beatles, Rolling Stones and then prog-rockers and heavy rock acts vied for chart-topping status, and ever since has been waning through the years of open systems and Unix, through Windows and on to the internet computing model of today. So be it: youth will always want to pursue the new and coming themes; the mainframe stands for mulish persistence and longevity but not novelty or the thrill of the new. And last, nobody is waving money under the noses of IT people to manage mainframes, so even those who seek no rewards other than pecuniary ones go somewhere else.

So there is a conundrum. We need to keep the machines but we need the bodies to run them. And the situation is exacerbated by the problem that many mainframe managers preferred not to share their wisdom but would rather keep it in an ivory tower located in their own brains and in the big systems in the datacentre.

The first and even second generation of mainframe admins are dying off -- quite literally in many cases. The data and processes stored inside these big systems are too valuable to write off and not easily transferred with satisfactory results. What is required is a formal skills handover, an application of knowledge management and a structured transition. But who will provide it?