[dylan604]

We have analog records from thousands of years ago, and we still have the technology to use them. I have digital formats from ~25 years ago that can no longer be read. I think the longevity speaks for itself.

Water and fire are equally destructive to digital or analog. Library of Congress has chosen analog vinyl for long term storage over any digital format. I think you're putting way too much faith in digital and totally discounting analog.

[adrian_b]

The majority of the very old records that still exist are those that have been carved in stone or bronze, or pressed in burned ceramic.

Besides these, there have been huge amounts of written records that have used more perishable materials, like plant leaves, plant bark, wood, papyrus, wax etc.

Almost all of those have been lost.

In the past, carving stone or bronze was hard work, so it was avoided, except for the most important records. Also today, there are technologies to make digital records that could survive for many millennia, but due to their high cost they are used even less than stone carving was used in the ancient world.

[stouset]

Sure, things carved into or painted onto stone have good longevity. That’s an insignificant minority compared to all the paper records that have been rendered unreadable over the same number of years.

Not to mention that accessing those records—finding them in the first place and searching through them—at any kind of scale is essentially completely prohibitive for most use-cases, many of them might as well not even exist until someone takes the effort to digitize them.

[squeaky-clean]

Can they actually no longer be read? Or would it just require a few hundred dollars spent on eBay?

[dylan604]

At this point, who knows??? None of the data is worth the few hundred dollars nor my time worth messing with it to be honest. I have a couple of Syquest disks that I keep around not so much for the data, but they're just kind of nostalgically cool to me.

[chmod775]

> We have analog records from thousands of years ago, and we still have the technology to use them. I have digital formats from ~25 years ago that can no longer be read.

I have physical records from 2 years ago that can no longer be read.

[stouset]

I explicitly called out that digital records having questionable archival properties doesn’t magically make physical records durable by default.

[jwells89]

It would make a lot of sense to develop some kind of digital storage medium that’s reasonably easy/cheap to read and write and is designed to endure centuries and hopefully millennia. Of course, this would need to come alongside a reader device which is similarly durable to ensure effectiveness (archives that can’t be read might as well not exist), which I suspect is the greater challenge.

[tbrownaw]

Pressed CDs are supposed to last a while. And I think there are archival versions of CD-Rs.

Or just go for something like microfiche (or even ordinary-scale engravings) with OCR-friendly fonts.

[dylan604]

except anything you can build today will be slow and tiny capacity tomorrow. the fact that anyone could suggest a universal format can be developed that will endure all of time where everyone continues to use it for ever after it is introduced is farcical like a fairy tale where unicorns exist.

[jwells89]

I would imagine that the express purpose for this format would be exclusively ultra-long-term archival and that it would not see much if any day to day usage, similar to tape drives now (but on a much longer timeline).

For this purpose speed and capacity aren't as important because we've had great information density for relatively small file sizes for a long time now. A ridiculous amount of information can be stored in just plain old text files, especially if compressed. For video and audio, recordings in even just 128k AAC and 720p h.264 video would be a wonderful thing to have 100+ years from now; anything better would be a cherry on top.

[mlhpdx]

Per the US Library of Congress, recommended formats:

https://www.loc.gov/preservation/resources/rfs/TOC.html

[necovek]

That's interesting: it focused on XML/SGML formats, but gives no mention of TeX (specifically LaTeX) that's pretty big in natural sciences and publications. It also seems trivial to reproduce with the base tools being free software and pretty small.

[pennomi]

Survivorship bias. I have no doubt there will be some subset of digital data that survives thousands of years. Probably Rick Astley honestly.

[Yawrehto]

>We have analog records from thousands of years ago, and we still have the technology to use them. I have digital formats from ~25 years ago that can no longer be read. I think the longevity speaks for itself.

We do have records from thousands of years ago, and the fact that we can read them is really cool! But, your argument that therefore analog is 'more reliable' is sketchy. It's true that we have many digital formats that can no longer be read today. But there a few things of note.

One, this isn't a fair comparison. We have analog records from thousands of years ago, yes, but we also have tons of missing ones. (Beowulf once had only one copy, and that was nearly destroyed in a fire in the 1700s. If someone hadn't been quick to rescue it, we could have no idea of Beowulf.) The vast majority is probably missing. A fairer comparison would be to invent a time machine and fast forward a few hundred years. If any file format from today is still readable, then that can be considered a success. Of course, absent a time machine, there's no real way to test this, but judging analog records only be survivors is flawed. (An incomplete list of books which we are aware of but don't have is at https://en.wikipedia.org/wiki/Lost_literary_work)

Two, water and fire might be equally destructive. But if, in theory, all of the print copies of the books in this collection [https://babel.hathitrust.org/cgi/mb?a=listis;c=158633464] get destroyed before digitization, then, so far as we can tell, they're gone. Forever. There's no way of recovering them. But if you can't access the file at HathiTrust anymore, for whatever reason (which might be more likely), there's a copy at the University of Illinois [https://digital.library.illinois.edu/items/bf8640e0-061b-013...] and at HathiTrust [https://catalog.hathitrust.org/Record/100008794]. And since HathiTrust has a mirror, in reality it would require three failing simultaneously.

Maybe a digital file is more likely to be unreadable or fail in some other way, but there are also many more copies. For a Project Gutenberg work to be inaccessible, all of the sites on this list [https://www.gutenberg.org/MIRRORS.ALL] must fail (in reality, it's less, as many are hosted by the same group; but it's still two continents, four countries, and eight providers) and possibly/probably others. For a rare book collection that isn't digitized, it takes one fire.