Preserving Data VirtuallyPreserving Data Virtually

For centuries, books have done a great job storing the collected knowledge of our civilization. Now, we're moving away from ink and paper to zeros and ones. Books stored digitally take up almost no space (hundreds of thousands of pages can fit on a CD-ROM), can be copied and distributed easily--and don't fade or crumble like paper.

But there's a major drawback to digital archives: While books printed 300 years ago work the same way as those bound today, data recorded just 20 years ago can be indecipherable on today's equipment. What use is a perfectly preserved Word file if a thousand years from now, nobody has a copy of Microsoft Office or a Windows machine to run it on?

It's a problem that's particularly acute for archivists at the Library of Congress, which has been working for years to transfer as much as possible of its massive collection to digital format. Last year, lawmakers allocated $100 million to examine the problem, and now some innovative solutions are coming to light.

The creation of a "universal virtual computer," software that would emulate a simple, logical machine, has been proposed by Raymond Lorie, a research fellow at IBM's Almaden Research Center in San Jose, Calif. When developers introduce a data format (like Word or Adobe's PDF format), they'd also produce instructions for the universal virtual computer, telling it how to decode the file. That means in the year 3000, librarians won't need to worry about how to read ancient file formats, because the machine will know how to do it for them.

"In the future, you'll only need to have an interpreter for the virtual machine," Lorie says. "For every new generation of machines, somebody will have to write that" interpretation.