Something I only figured out late: software doesn't age like a book. A book yellows, but anyone who can read can still read it. A 1994 program on a perfectly intact floppy disk can be, for all practical purposes, unreadable. The machine is gone, the operating system, the monitor it depended on, sometimes even the mental habits of the people who used it. Software preservation is a permanent translation problem, and almost everyone treats it as a storage problem. That's why the Internet Archive interests me less as a library and more as the only place that took this translation seriously. In 2015 they got MS-DOS games running in the browser, a little over two thousand titles in the first batch, using a version of DOSBox compiled to JavaScript through Emscripten. It looks like a trick. It isn't. You no longer have to install anything, configure anything, work out why the sound isn't working. You click and Prince of Persia boots. I've watched people who have never opened a command line in their lives play Commander Keen in a tab, and I think that does more for preservation than a hundred papers on digital archives, because a memory nobody can experience dies anyway, just more slowly. Then came Flash. When Adobe killed it at the end of 2020, there was a strange end of year in which thousands of games, animations and entire advertising-agency sites lost the ground under them overnight. The Archive took Ruffle, the emulator written in Rust, and integrated it. Honestly, it was one of the few times I've seen an institution react in time to a death announced years in advance. The other times it was always late. Money and law are the real bone. To preserve software, you have to copy it. To copy it, you often have to get around a protection. Getting around a protection, in the United States, runs into section 1201 of the DMCA, and the only crack in the wall is the set of exemptions the Copyright Office renews every three years, in a process that looks designed by someone who has never seen a floppy disk. The most recent round, in 2024, once again left out remote access to games preserved by libraries and museums. You can keep them. You can, in certain cases, study them on site. Letting a person in Braga play, from home, a title nobody has sold for twenty years is still risky territory. The number that stuck with me most, and it isn't from the Archive, comes from the Video Game History Foundation: a 2023 study, done with the Software Preservation Network, found that around 87% of classic games released in the US are commercially unavailable. Nearly nine in ten. Anyone who says the market is handling this is talking about something that doesn't exist. The market sells what sells well, and the rest waits for a company to fold, a server to shut down, a programmer to retire. Now the uncomfortable part. The Archive didn't help its own cause with what happened in 2020. The National Emergency Library, which lifted the lending limits on digitized books during the pandemic, was the spark for the lawsuit from Hachette and other publishers. They lost in the first instance in 2023 and lost again on appeal in 2024. I'm not sure the ruling was unjust, and I've changed my mind about it twice. But I know the second-order effect was terrible for everything else. Every defeat of a book project weighs on a software project, on music from 78 rpm discs, on web pages, because in the eyes of a judge and a legal department all of this is "the Internet Archive", one single thing, with a target painted on it. And the target kept getting hit. The record labels' lawsuits over the Great 78 Project, which digitized old shellac recordings, dragged on until they were settled out of court. In October 2024 there was also that DDoS attack and the data breach that exposed around 31 million accounts. I have no proof of any link between these things and I'm not going to invent one. I'm only saying that a nonprofit with a library's budget, being attacked by lawyers and by people with botnets at the same time, is not a sustainable situation. It shouldn't need saying. I tested something small recently, out of curiosity. I looked for a shareware program I vaguely remembered from the time I used to assemble computers in an uncle's garage, and it was there. The file, the opening screen, the registration notice I never paid for. It took about 4.2 seconds to start. I stood there longer than I should have, without wanting to admit why. By the way, my own track record here is miserable. Once, out of laziness on a shared server, I lost the only backup of an old project and never recovered it, and I still get irritated when someone tells me "the cloud keeps everything." It doesn't. It keeps it while someone pays the bill. Some people argue that emulation is an illusion, that an arcade game running in a browser isn't the game, because the cabinet is missing, the cigarette smell, the latency of a CRT. They're right about part of it. But it's a purist's argument that, in practice, almost always serves as a reason to do nothing. I'd take a living imperfect copy over a perfect original nobody can open. What really worries me is depending on a single place. If the Archive falls, and it has been closer to falling than we like to admit, there is no plan B at the same scale. There's Software Heritage in France, which archives source code and has a different purpose, there are academic initiatives, there are stubborn collectors with hard drives in their basements. All useful. None equivalent. I don't know how this ends. Maybe with a better law, maybe with an ugly settlement, maybe with a shutdown date nobody announces. What I do know is that the next time a format dies, and it will, someone had better already have the translation tool ready.