This USB flash drive can only store 8KB of data, but will last you 200 years

schizoidman@lemmy.ml to Technology@lemmy.ml – 223 points –
This USB flash drive can only store 8KB of data, but will last you 200 years
tomshardware.com
71

The year is 2245. The heirs finally locate a working, antique reader that can handle the ancient USB key, hoping to find great-great-grandpa's crypto-wallet or the pin-code to a long-lost Maltese bank account.

Instead, they find a 4-bit, VGA-quality scan of Miss October.

An actual book stores more data than that and for longer. At that point, why not just etch the data onto a metal plate or something? 8K is only a few pages of text at 12pt. It could easily fit onto two sides of a small-ish metal plate, etched in 8pt or so, and it would last, potentially, for millennia.

I think the idea is to improve upon this tech so the capacity would become larger.

It’s FRAM, which has been around for ages. The problem is its prohibitive cost— hence the 8k.

This can be rewritten many, many, many times.

What’s the practical benefit of that? If the point is long-term storage, rewriting isn’t a priority (or possibly even a need). And this isn’t designed for capacity.

It's so I can exchange fart jokes with my great great great great grandson via a magic USB port a la The Notebook, assuming that's how it works, idk, never actually seen the movie.

I usually update my backups with new TOTP or other cryptographic keys every few months. Sometimes every few days (when I generate new keys).

Why even invent the car when horse so much faster?

idk, its much more resistant to floods.

Than an etched metal plate?

Yes.

USB drives don’t mix well with water

Why do you say that? I've put several through a washing machine and gone swimming with mine. Electronic usually are fine with water. Batteries are not.

Just let it dry out before passing a charge through it, and it works fine. Not so easy with paper.

Not so easy with paper.

I had suggested a metal plate, not paper.

So your 8KB of data will last forever, but what about the firmware required to access it running on flash?

In 200 years, AI will hack it for you, but you’ll need a dozen antique dongles to get from USB-Z to A.

it does say it has a built-in serial console and raspberry pi

And the firmware inside that rp2040 is stored on plain old flash memory. So while the data may still be on the memory chip, the controller chip dies at just the same pace than every other usb drive - and then you can't access it.

  1. Print out 8KB on high quality paper.
  2. Store in good environment...

How do I read that data back

With your eyes

But 100 years we’ll all be mole people without eyes!

My brain doesn't have the decryption key. I'm no man in the middle.

It's the same problem with a drive like this, or any long term archive, you either store the data unencrypted and rely on physical security, or make sure you store the encryption key and algorithm for the same length of time, in which case you still need the physical security to protect that instead. In both cases you need to make sure you preserve a means to read the data back and details of the format its in so you can actually use it later.

Paper is actually a pretty good way of storing a moderate amount of data long term. Stored correctly it's unlikely to physically degrade, the data is unlikely to suffer bitrot and it can be read back by anything that can make an image in the visible spectrum. That means you can read it, or take a photo and use OCR to convert it into whatever format is current when the data is needed.

just print something like a QR code in absurd resolution and read it in a document scanner, a single sheet of A4 should be able to fit quite a lot of data.

I was curious, so I looked it up and it seems that around 3KB is the max for a single 177x177 code (though I imagine this is a "soft" limit?). With 600DPI being common for laser printers, a DPI-limited 3KB would be well under 1cm x 1cm. My hunch is that this wouldn't be super reliable (DPI limit not necessarily the resolution of the printer?), but I'd be curious to see what the usable QR density actually is. But yeah...a few QR codes should do the trick!

also QR codes have EXTREME data redundancy, you can cover like half of a QR code and it'll still work.

In our scenario we don't need much redundancy, since the paper will be in controlled conditions and shouldn't degrade, and we'll make damn sure to scan the entire thing without crap obscuring it.

We also don't need all the tracking features, all we need is a marking in one corner so we know what way to put it into the scanner.

All this taken together should result in a data density that is actually realistically useful with just one side of a single A4 sheet.

Imagine storing a digital photo on a piece of paper, and needing to scan it to reproduce the photo.. someone needs to do this!

  • Take picture with digital camera.
  • Store as jpg
  • Convert jpeg to base64 string
  • Compress the string in a .zip split into 2.5KB chunks
  • Encode the .zips as a base64 strings
  • Render a QR code for each string.
  • Print out all QR codes on a sheet
  • Store in family photo album.

Most of that could be like 10 lines of python...

I'm going to get one for my favicon.

What's the favicon?

I would be surprised if you couldn't get 8KB for 200 years out of standard flash simply by extreme duplication --- 8GB/8KB means a million copies on one (very small by today's standards!) drive.

Or is the failure mechanism something other than bitrot?

Kinda funny, I was just writing about archival media this morning. Verbatim makes DVDs & Blue Rays that last ~100 years, and M-DISC makes ones that'll last ~1000 years. And the Verbatim Blu Rays run ~$0.036 per gig.

Where can I read about the testing that goes into these claims?

I'm far from an expert, but anything on standards JIS X6257 / ISO 18630 would probably be a good start. It's an open standard for 100+ year discs.

Otherwise probably best to look into accelerated aging studies. For technology that's less that 100 yrs old to claim 100 or 1000 is a bit uncertain but accelerated aging is probably the closest to a best guess. I recall skimming over a third party lab saying Verbatim gold foil archival DVDs were estimated to last 30-120 years depending on storage methods and luck, but never saved the link.

I'm sure storage environment matters though?

Oh yeah, stick it in the sun or a damp box and either will probably be bad in weeks instead of decades or centuries. But supposedly they'll meet those lifespans good at room temp

Will the chip actually last that long though? I would have expected a ceramic package with gold plated leads, not a plastic SOP-8.

Somewhat better than this useless USB thingy (from Temu?)

https://futurism.com/memory-that-lasts-forever-new-quartz-coin-can-store-360tb-of-data-for-14-billion-years

Summary by Andisearch

Researchers have developed a new quartz coin that can store 360TB of data for 14 billion years. This is a significant improvement from the previous quartz glass storage, which could only store data for 300 million years. The technique uses femtosecond laser pulses to write data in the 3D structure of quartz at the nanoscale. This makes it possible to store the whole of human history in a small coin-sized device. The storage system is also very durable, able to withstand high temperatures. This technology could potentially serve as a means of archiving important information for future generations or even extraterrestrial beings.

You mean my 2GB Kingston that I bought in 2007, that I rarely use anymore won’t last me 200 years? Damn…

The other flash chip storing program code for the rp2040 will decay before then making the longevity marketing dumb

Kind of hard to return this after 195 years if it fails.

what about physical damage. or emp or something. I feel like that will be a problem well before 200 years.

Mold is actually the biggest concern with the most popular archival format LTO. EMPs aren't that much of a concern. Bit flips and bit rot are your main concerns traditionally when using flash for archival storage. It's recommended if you go the flash route to keep your array hot (ie powered on) and use a file system with data scrubbing capabilities such as ZFS.

The more you read is just gets wilder and wilder.

I'm admittedly interested now.