When I blogged last week on the end, at last, of further Microsoft updates to Win 9x, the spectrum of reactions ranged from resignation ("We'll keep on keeping on with 98SE until the last electron dies") to contempt ("If you use a Win 98 machine on legacy hardware, you are slow and inefficient"). Most noteworthy, it seemed to me, were these four points that appeared in various combinations and intensities:

• Legacy applications often lack upgrade paths (no source code, no budget for rewrites) to new platforms
• Lab or industrial applications may depend on specialized hardware with no access to updated drivers
• Fully amortized hardware that runs Win 98 quite well may be quite inadequate for current Windows versions
• Validated systems in regulated industries entail forbidding costs of platform migration

I said these points were noteworthy because, although not very surprising, they come from the trenches where people are justifying IT investments and trying to meet their numbers—not from the bright but often noisy plasma cloud of the technology community, where technical improvement creates its own demand.

The people whose comments I'm citing represent the second full generation of PC application, starting that clock in late 1981 with the debut of the IBM PC line. They're no more impressed by the technology in the box than my teenage sons are by cellular phones. It's just a unit of function, as versatile as duck tape and just about as sexy—which is to say, not very sexy at all.

One of the most specific suggestions represented, in effect, the encapsulation of an existing system: Put Firefox and Eudora on a Win 98SE machine, and never run IE or Windows Update. It reminded me of case study interviews that I did in 1988, one of which elicited a comment about an old IBM System/34 in the corner: Given the dearth of new-hire coders with any knowledge of RPG, I was told, that box might as well be a single specialized component that they'd just keep on running until it died.

It's kind of shocking to count the ways that an application only a few years old might get trapped in technology's past. It might directly address the serial or parallel port. It might be written in an early version of Microsoft's Visual BASIC or dBase II. For any number of reasons, a perfectly good piece of your business technology portfolio might suddenly be recognized as having lost its connection to the present, let alone the future, of the systems you might want to build.

It's interesting when a living fossil like the Coelacanth emerges from the depths of the ocean, but it's no fun at all when the fossil shows up in your data center—or perhaps on a departmental server at the company you've just acquired, or on the PC of a key employee who's shortly to retire. Episodes like the Y2K codequake should have taught us to maintain inventories, not merely of applications, but of the bundles of associated platform components and skills required to keep them viable. Perhaps six years of ever-tightening IT budgets have squeezed the vigor out of that resolve; perhaps it's time to renew it.

Tell me about the fossils you've excavated lately at peter_coffee@ziffdavis.com.

Click here for an archive of Peter Coffee's columns.

This article was originally published on eWEEK.com.