/Technology

Before decent CD-R burners came along that didn't fail every other attempt, backing up data was basically an act of faith. People didn't have a "backup strategy" in the sense we use the term now, with a 3-2-1 rule and automatic cloud sync running in the background. They had one method, usually just one, and hoped it would hold up when the moment came. Punch cards opened the chapter, back in the 50s and 60s, inherited from Jacquard looms and adapted for early commercial computers. Each card held a line of code or a single record, and whole programs turned into box after box of cardboard full of little rectangular holes. A typo meant repunching the entire card. Dropping the box and scrambling the order was, without exaggeration, one of the most common nightmares for anyone working on a mainframe back then — there was no undo, just the job of putting everything back in order by hand, checking it against notes. Once reel-to-reel magnetic tape entered the picture, it solved part of the volume problem but brought new ones. Data centers adopted a rotation scheme that still lends its name to modern practice: grandfather-father-son. Today's tape became yesterday's, which became the day before's, and so on, a hierarchy that guaranteed a few generations of recovery without needing infinite tape. Except tape stretches, loses alignment, and after a few years sitting in a poorly ventilated cabinet the oxide starts flaking off the plastic base. Some people went to restore a ten-year-old backup and found the tape basically crumbling as it ran past the read head. The home computing side of the story is maybe the most charming, and also the most frustrating for anyone who actually lived it. Ordinary cassette tape, the same kind that played music on a Walkman, became a storage medium for early micros — the Commodore 64, the ZX Spectrum, even IBM PCs shipped with a cassette port almost nobody remembers existed. Commodore's Datasette, for instance, ran at around 300 bits per second, converting data into audio tones through FSK encoding, essentially a very slow modem recording onto ordinary magnetic tape. In practice that worked out to something like 60 to 70 bytes per second, meaning a 60 KB program could take fifteen minutes to load — and that's assuming you didn't get a LOAD ERROR halfway through and have to rewind and try again. Fastloaders could multiply that speed by ten, but nobody would call any of it reliable. A tape left too close to a speaker, or just misaligned from the factory, and there went the only copy of a school project. Floppy disks brought random access, which was already a massive step up from having to run the whole tape looking for a file. But capacity grew slower than demand: 360 KB on a 5.25-inch disk, then 1.44 MB on the 3.5-inch, and that stayed the de facto standard for way longer than it should have, considering how much bigger graphics files and documents kept getting. Storing a design project across a stack of numbered floppies, hoping none of them failed mid-sequence, was routine. And floppies fail easily — a stray magnetic field, humidity, or just the mechanical wear of the read head scraping the same track hundreds of times. Trying to solve that capacity bottleneck is what gave us the larger removable disks, things like the Bernoulli Box and, more famously, Iomega's Zip Drive, which became a craze in the mid-90s offering 100 MB on media about the size of a thick floppy. For a while it looked like the definitive bridge between the dying floppy and the age of optical discs and USB sticks. Except the Zip Drive carried a serious flaw: the infamous repeated clicking, known as the "click of death." The mechanism would try to find track zero on the disk, and when it failed, the head would bump against its movement stops and end up misaligned — and the worst part was that a "clicked" disk could damage other disks when inserted into a different drive, which gave the failure a reputation almost like a contagious disease in forums at the time. Iomega denied the problem for a good while before eventually facing a class action lawsuit, settled in 2001. For a lot of people who'd trusted their entire backup to a single Zip Drive, the lesson came too late. Companies with bigger budgets used DAT tape or dedicated magnetic tape drives, usually running overnight, in a manual process of swapping the cartridge and physically driving it to another building, sometimes another city, purely for fire safety. That's not a figure of speech: entire services existed just to pick up backup tapes and store them in a fireproof vault far from the original data center. The term "sneakernet" describes exactly this — the most reliable data transport method of the era was a person walking with the media tucked under their arm. What ties all of this together, from punch cards to the Zip Drive, is the lack of decent error correction and the cost per megabyte that limited how many copies of anything people actually made. Backup, in practice, was often just one copy — not a redundant system, just a second place where the data existed, one that could fail just as easily as the first. The idea of keeping three copies, on two different media, with one stored off-site, something that's basic advice in any backup tutorial today, was a luxury for large data centers, not something an average user could pull off. Funny enough, the problems never really went away, they just changed shape. Bit rot still happens on SSDs. Cloud providers still go down, still lose data, still suffer silent failures that only show up right when you need to restore something. The difference is that today it's easy to forget how physically demanding it used to be to keep a copy safe, since we've outsourced almost all of it to a sync button that runs on its own. Worth remembering where we came from every once in a while, if only to avoid repeating the classic mistake of keeping everything in one place.