50 years ago and all those letters that were burnt or composted because no one bothered to keep the sentimental letters of soldiers in a world war.
20 years later and all those millions of photos that were on Flickr but no one bothered to upgrade.
100 years later and all those GPS logs from taxi cabs that were logged by firms are lost as no one thought they would be interesting or useful.
600 years later and these floppy disks with meeting minutes on them about an early internet organization are lost but might well be worth more than they are now.
Roll-over to new tech is safe. The content of discussion is inherently reliable due to its digital nature.
Keeping multiple copies is necessary practice and the long-time trend is that the next copy you produce has no greater cost than the previous copy (if you wait years, it will be significantly cheaper).
"Just". Have fun justifying the cost of re-encoding and re-archiving all of the laws of the UK every few years. The advantage of vellum is that once it's written, that's it. You don't need to go back and re-encode and upgrade to the latest storage technology. It's there, and it'll last literally a thousand years if it's stored in a reasonably dry, cold, and dark location.
But you need to avoid the mistakes that people have already made: using some weird format that dies, and then having all your data locked away in a semi-readable file.
Some formats that are no longer usable in general, such as punch cards, were obsoleted consciously. And you might struggle to find a LaserDisc reader. Otherwise, we should be fine. Or am I missing something?
"The third disk here is the only known copy of the original design documents for Loom. It's an 800K Macintosh floppy employing a proprietary format readable only by a vintage Macintosh drive (thank you Apple). In preparing for this lecture, I obtained a dusty old Mac which had not been turned on since 1997. After reseating all the cables and boards I got it to boot and determined that the files on that disk are still intact and still fully readable. A few moments after making this happy discovery however, and before I could actually retrieve those files, the hard drive in the Mac crashed and died permanently. I'll recover the files eventually, but for now you'll have to rely on my failing memory for how Loom was conceived."
[1]: http://ludix.com/moriarty/loom.html (I transcribed the passage above from the video)
So imagine you have generations of digital files, converted from format to format to format. Why? Because software engineers are assholes who keep inventing new formats, for stupid goddamn reasons. A hundred generations from now, you'll be looking at copies of copies of copies which aren't just checksumable bit-wise copies but transcriptions, with transcription errors. If our descendants are lucky, all of those different versions will be preserved as well, but that just means that if you notice a possible transcription error, you'll have the opportunity to dig into two hundred year old character and file encodings to try to figure out what the original text was.
Which is not that different of a situation from trying to guess whether a scribe three copies ago misread the scribe four copies ago's atrocious f for a t.
"During the 1960s, the punched card was gradually replaced as the primary means for data storage by magnetic tape, as better, more capable computers became available. ... [P]unched cards were still commonly used for data entry and programming until the mid-1980s when the combination of lower cost magnetic disk storage ... made punched cards obsolete ..."[1]
Yes, I listed wikipedia as a primary source, deal with it.
I'm not sure if I'll ever be able to open them. I don't even have a way of getting them off the floppy.
But yes, I use it, both professionally and personally. Legacy MUMPS is not pleasant -- the VA's code looks like someone left their telegraph out in the rain -- but the language is astonishingly good at string-handling, mostly due to certain constructs that make it very hard to translate into normal languages.
Edit: On further consideration, I think we have to regard digital media and formats like a form of encryption (a simile, I don't mean this literally) which gets stronger over time. Time will bury extant evidence of what is contained on the media and how it is stored. Why should not make the information as explicit (plain-text) as possible? People in 800 years will still have to spend their efforts decoding early 21st century English legalese.
Compare and contrast: https://en.wikipedia.org/wiki/Domesday_Book#Custodial_histor...
Ah, the classic sign of someone about to say something rash about technology.
On the other hand, storing data for an active/continuously used archive doesn't seem so hard. You just keep the data digitally on some media and every few years you upgrade the tech to whatever technology and formats are "current" before the old format and tech is completely forgotten. Keeping digital data around for a decade is pretty easy. If you do that 100 times in a row you have kept it for a thousand years and you never had a file format or a storage media older than a decade. You could "forget" to upgrade the archive, but you could also "forget" to scribble on the calfs some time.