A little off topic but:

The 1960 US Census was the 1st computerized census, all the data was stored on magnetic tape, IBM discontinued the tape drive in 1965.

A friend is in the computer network installing business, he was at a client doing some
work and saw a pallet of hard drives and a computer marked BACKUP on the loading dock, he asked about it and was told it was the year end off site backup of the company data.

They put all the data on the hard drives and shipped it off to some salt mine for storage, they included a computer so that they were sure to have hardware to read the hard drives.

Smart. That is a big issue. Like James was talking about with a Win95 box that should've been able to read a contemporary USB memory stick but couldn't. And hardware makers won't help, they want to force you to buy new hardware. So burning a ZIP disk of your beloved data and slapping it in a drawer for 20 years, is highly unreliable.

The only reliable way is to frequently "bucket brigade" your backup data from format to format as they keep coming out. Consider the fact that disk capacity seems to double every 2 years, there's a lot of sense to just keeping everything from day 1 of your digital life on your current hard drive and part of your active backup regimen. In other words drag all your digital data along forever, and keep updating how it's backed up so it's current. 10 years ago you used backup tape or CD-Rs, now you use the cloud and cheap hard drives. Whatever.

However there's a second key problem. "Format rot". Your data file survived the decades, but it uses a binary data format which was proprietary to an application, and no application that can run on a modern system can read it. So you had to keep bucket brigading your files forward as formats and dominant applications evolved... from Visicalc to 1-2-3 to Excel. But modern Excel won't read Excel 1.5. There's format rot.

This is a problem that's been known since the 60s, when GML was invented, followed by SGML, followed by XML -- as attempts to come up with a universal, self-documenting, internal file format that would give future computer archaeologists a shot at decoding the data. Applications didn't use them because translation was slow and took more disk and memory, and that was a competitive issue. None of that is an issue nowadays, and applications are starting to adapt to these new formats. For instance the new .xlsx Excel format is XML in gzip, both rock solid standards.

The takeaway from this is -- if you have data in applications which are proprietary, occasionally export them into a standards based format as part of your backup regimen.