And now for something COMPLETELY different:
Linux booting *inside* a PDF file. (Uses embedded JS to run a RISCV emulator and boot Linux to a command line prompt):
https://github.com/ading2210/linuxpdf?tab=readme-ov-file
https://www.theregister.com/2025/02/16/dev_linux_pdf/?td=rt-3a
@cstross OMG. There's a whole new rabbit hole for me to jump down later!
The obvious next step is to run Doom.
Whoops, never mind: they've got that covered, too.
@cstross Wow! This is a great example of the power of virtual platforms, #emulators, and also disciplined state control in a programming language. The fact that you can run this inside a #pdf is also instructive — yet another attack vector.
Also, I’ll bet there are many packages out there to do cross-compilation from #RISC-V to other machines. Remind me never to use Adobe Acrobat.
@cstross Yet another reason to add to the list of why friends don’t let friends open PDFs…
@cstross Yet another reason to add to the list of why friends don’t let friends open PDFs…
Crazy impressive and absolutely mind boggling though. 😅
@cstross When I last worked with laser printers, there was a PostScript program that would run Doom - about 15 yrs ago. We never had the time to actually test it, but had to do some extra work to disable the ability to do arbitrary IO.
That said, amazing job and I'm wondering if I should replace my PC with a good printer!
@cstross This reminds me of the guy who stuck an LLM inside a font. Actually, there is more than one guy who has done this.
https://www.youtube.com/watch?v=Y65FRxE7uMc
(this explains how he uses the LLM to rewrite your text using synonyms so as to achieve pleasing justified text without use of hyphenation).
https://fuglede.github.io/llama.ttf/
(this does more conventional predictive text stuff as you type.
I'm not sure what happens when you typeset your PDF-containing-Linux in a typeface-containing-an-LLM, but I'm pretty sure I don't want to be there when it happens.
@cstross
I was gonna joke that they need to run Doom but i guess that's what this is actually about 🙃
@cstross @bsdphk Computers were a mistake. That ANYONE thought, intentionally or accidentally, that the ability to run this much code inside of a document deserves to be shunned back to the Stone Ages. This is why infosec is an impossible game to win. For every one person fighting the good fight, there are a hundred shitty programmers writing vulnerabilities like this and a thousand scammers waiting to use them.
Next, a #kubernetes cluster in a PDF file...
@cstross truly this is the year of linux on the desktop 🤣
@grob Doom already happened! https://github.com/ading2210/doompdf
@jgeorge @bsdphk Blame Adobe for adding JavaScript to PDF, back in the days when everyone wanted to slather world wide web special sauce on their products—like LLMs and "artificial intelligence" today or "crypto/blockchain" a decade ago.
It's a repeating dark pattern, dictated mostly by managers desperate to make Line Go Up by being latest-buzzword-compatible.
One of the events in computing that I have yet to see credibly researched and documented is what I call "The deluge":
How much did "The IT industry" grow during the 1990'ies in terms of persons, and what skills did they come in with ?
My guess is a factor of more than 1000, (ie: 100000+%) and that none of those people (-epsilon) had any professional IT experience, and knew nothing about the skill&craft of IT.
(Add the "Make Money Fast" incentive and...)
@bsdphk @jgeorge You can almost certainly chart it by graphing cumulative sales of personal computers, minus machines retired as broken/obsolete, which lags by a time period … and it was undergoing exponential growth throughout this period, so the retired cohort lagged new installs by a huge distance.
Computer use in business implies support personnel numbers (IIRC in the 90s it was 1 PC tech support per 40 installed base).
Since 2007 include smartphones/tablets as well. Different ratio, OFC.
That's pretty much how I arrived at "a factor 1000".
But having somebody who know what they're doing, do the math with actual data would be better :-)
@Sweetshark @jgeorge @bsdphk And Turing-completeness is an astonishingly low bar. At least, it's astonishing to folks who haven't studied CS. (sendmail.cf? Turing-complete. Magic The Gathering? Turing-complete. And so on.)
