RE: https://tech.lgbt/@nina_kali_nina/117184991521312864

There's no new "life with NetBSD" report due to my semi ongoing health issues, so I'm going to boost the previous one.

I'm thinking, when I have time and energy, about a personal project related to surviving in the era of collapse of personal computing. Likely it'll be just a blog about making things cozy for me and my partner, where we would be exploring topics like:

• building offline-only Debian workstation based off old-stable, and potentially fetching some security fixes from upstream
• making NetBSD a simple desktop OS with minimum slop/no AI-tainted packages coming from upstream
• some minimal risk evaluation for software we use in terms of "will it bitrot, will it be taken over by fash"
• exploring alternative OSes more: not just BSD, but Haiku, Redox, FreeDOS
• collection of links to cool human-made projects
• etc etc

This is a gigantic project that I feel like I have to do for my own sanity. It'll take a long time to even get started.

RT: https://tech.lgbt/users/nina_kali_nina/statuses/117184991521312864

While I'm still feeling meh, I want to at least call your attention to some projects that are worth highlighting:

cproc C compiler
https://sr.ht/~mcf/cproc/

Did you know that GCC allows fully LLM generated tests and some amount of LLM code in codegen? Did you know that LLVM allows even more? That tinycc main branch is free for all with vibe code? That author of chibicc is now full on in vibe coding, and even vintage pcc, watcom and tack might be endangered to slop? It's all very, very sad and scary.

Cproc is the only C compiler that I know of that is explicitly not allowing LLM output in its code. It is reasonably mature, as it can compile a Linux distribution called oasis, and it is fairly portable - e.g. I had very little problems of building simple programs with it on NetBSD amd64 - but it is not a drop in replacement for GCC.

(I forgot 9front's compiler, it's fine too, but it's not the C flavour I know)

🧵 Continues

@nina_kali_nina there is also https://github.com/fuhsnn/slimcc which is a fork of chibicc, but is now basically a different compiler after the years

Software projects you probably should know about, continues

• servo browser engine https://servo.org
• Sea Monkey https://www.seamonkey-project.org
• PaleMoon https://palemoon.org
• Dillo https://dillo-browser.org
• Netsurf www.netsurf-browser.org

Chromium and Firefox rely on LLM generated code extensively. This means Firefox forks and Chromium forks are AI-tainted. WebKit, too, has Claude and Codex generated code in it. This means ways of browsing the web, the modern web, without slop, are critically endangered. If you want to reduce the harms of LLM usage, even moving to a Firefox fork would be a net positive thing, but our next best bets for slop-free web browsing are either Servo (warning: Rust and some dependencies needed for it are somewhat AI tainted) or forks of old Firefox, I think. And alternative js-free browsers.

I'm really, really sorry, but I think modern web might not make it. Smolweb? Sure. Let's use smolweb more

🧵 Continues

@SRAZKVT this is great news! I hope the project will have more explicit anti-AI policy. I am afraid we're at the point where projects that didn't have one might end up getting a policy from Linux kernel, systemd, Debian, Alpine and other projects embracing slop.

@nina_kali_nina so tor-browser is basically fscked as well? Well I can run something else through whonix, but still. tor-browser back.tracked to esr after trying main for a while, anyways, yeah good info, nina

@nina_kali_nina > and even vintage pcc, watcom and tack might be endangered to slop

How? I'm confused How do you sloppify something like pcc that hasn't been updated in 10 or 20 years?

@toots IIRC they're doing a lot of work to avoid slop, but the next Firefox ESR is full of LLM authored code and it's not looking great tbh :(

@vaporeon_ right?!

@nina_kali_nina yea

but honestly, i don't think c compilers is an area where there is much of an actual worry, they aren't impossibly simple that any student can make their own, but they are simple enough that anyone with time could make their own, and so jumping from one to another isn't a big ask so long as you keep to only the standard (and maybe only very common extensions)

@nina_kali_nina
Worth mentioning that cproc is a front end to QBE which is also used by Hare.

QBE is a compiler backend that aims to provide 70% of the performance of industrial optimizing compilers in 10% of the code.

More front ends for QBE would be nice, assuming the slop machines don't get to it.
(and even if they do, it should be much easier to handle a fork than gcc/llvm)

@nina_kali_nina FreeDOS has slop now, fyi

@toots @nina_kali_nina TOR itself is just a protocol. Those responsible for TOR tooling and infrastructure don't have the means to make a browser. Heck, all browser specifications and standards are driven by the aforementioned "tainted" browser companies, so 🤷.

@mayke @toots by the by, I think Servo or Dillo or Netsurf shouldn't have any issues with running over Tor protocol. Tor Browser, thought, is in a sad place to be.

@nina_kali_nina @toots Ya. I also doubt the maintainers of the TOR software will go out of their way to keep the project entirely free of AI...to be fair they're likely hugely understaffed and underfunded to begin with.

@kevin I've written about it previously; long story short, they have one of the strongest anti-AI policies among all operating systems out there. They have no contributions from LLMs to the code (only to the text/translations), and they said they're committed to keep it this way. They do allow LLM code reviews, which I find regrettable, but all and all, if Linux had a policy like that, I'd be much happier than I am now.

See more in the thread: https://nerdculture.de/@boeckmann/117260336193862046

@nina_kali_nina @mayke hmm tor browser should really be considered as one of the Flagship utilities of the tor project, yeah I have tons of other options, being on a technical level, but it sucks for people who transparently trust the tor project, I hope they'll sharpen up then, and not mislead random people into false securities.

@nina_kali_nina I stand corrected! I read this somewhere and never looked into it, mea culpa.

@nina_kali_nina@tech.lgbt cuikc kinda cool aswell :3

@freyalikesgirls looks promising! I hope they'll not succumb to slop

@nina_kali_nina i'm sort of looking forward to the cyberpunk irl hardware aspect of this where we're all meeting up someplace wearing hoodies and flipping open our briefcase sized bootleg laptops made of interesting and scrounged together parts.

@nina_kali_nina@tech.lgbt wait tcc is tainted?:(

@freyalikesgirls last stable is fine, but the main branch is not protected and, as I said, it's free for all 😖

@snosrapkungfu this might end up more real than you think. Old hardware (pre-win11) is getting bulk disassembled as we speak, with RAM and SSD salvaged and everything else used for metal extraction. I've seen mountains over mountains over mountains of dead computers and it makes me want to cry. Microsoft did an incredibly harmful thing by artificially obsoleting whole generations of otherwise fine hardware. Computers that are pre amd64 are too slow to be used for modern software, and are rare and expensive due to collector value. Millions of mass produced Intel core, atom and AMD64 computers are going through a uh genocide? How do you call a process of destroying physical artifacts of a specific type, at scale?

@nina_kali_nina@tech.lgbt Tbf last release tcc just makes sense its not major

@freyalikesgirls but at this pace one might want to just use an old GCC then. I think BSDs might be forced to do just that.

@nina_kali_nina This is looking like a very interesting thread to read, so I'll have to look deeper with more time :-)

Do you think the project would/could also have a mailing list, a newsgroup and/or an IRC channel? Or are these less welcome ideas right now?

@nina_kali_nina oh dang seamonkey is still around :O and gentoo has a package for it :O

did they just keep going with whatever was current at the time the project was established, or do they use firefox's browser engine now days?

@nina_kali_nina netsurf is great but realistic modern browsing requires a minimal amount of JavaScript :(

I wish I could offer more advice and guidance on living without AI-tainted or anti-humanist/fascist software. I think if you're worried about it, and have energy and time, you should try keeping your existing software and hardware going for as long as it is feasible. If updates stop coming, consider moving things to be offline-only. Research repairable alternatives. Learn about emulation. Choose safer options, and you probably will find No-AI list as helpful as I did:
https://noai.starlightnet.work/list.html

Additionally you might want to be aware of Open-Slopware list, though it keeps the projects in the list to a very high standard, including things I personally find forgivable (e.g. MartyPC is in the list): https://codeberg.org/ethical-foss/open-slopware

Finally, if you want my opinion on Linux: I'm sorry but I think it's mostly done for, for my purposes. I don't think I'll find it fully ethical to use after LTS support for the late 5.x/early 6.x kernel drops. And don't get me started on systemd, python and other things that embrace AI.

🧵 end

@nina_kali_nina Your endeavor, including the sharing of it that you're doing, is very valuable for people with less expertise in these fields but a wish to resist ai as well. So thank you.

p.s. as I'm saying, I'm going to research things more, write about them more, as much as my health and time allow, but sadly there's no solution to the poly-crisis that'd fit all. I wish NetBSD or OpenBSD could satisfy everyone's needs - but they can't, at the moment, be as simple to use as Linux, and for many, moving to, say, NetBSD could feel like moving to a foreign country without most of your possessions and without your family and support network. The friends I made when I arrived to the BSD land are great, but I still feel like I've lost my $HOME. And I don't feel like my answers will work for everyone. So I won't tell you to do exactly what I did.

Instead, I think we should experiment more. Embrace permacomputing more. Embrace smaller communities. And among those, we might find new, cozy homes. Maybe Haiku will be great for you. Maybe you'll find 9front exciting. Maybe you'll enjoy writing in tal for uxn. Survive. ❤️‍🩹🫂 Stay sane. Look after each other. Stay kind to humans.

@f4grx In a way, it's not minimal anymore. Major sites that work just fine in Netsurf require JavaScript because of Cloudflare front. And even very advanced browsers are struggling with it. Therefore, RIP, www

@aeva they are a hard fork, and so is Palemoon+Basilisk. I don't know if they're great as communities, and I don't know how secure and up to date they are, but they're still supported

@nina_kali_nina imho you are very brave. I am still appalled and finding myself stuck at the acceptance phase that most everything that I was keen on learning and using at a deeper level of knowledge (e.g. Linux, all of the major desktop environments, python) has been/is in the process of being destroyed. I have very very little energy and feel terrified of making such extensive changes as changing to BSD, learning yet another language, finding and learning replacements for most of the programs i use.

Might as well make my 2008 iMac full of pirated software/abandonware my main machine again.

@me it is Dennou Coil, an anime which dystopian future feels nicer than dystopian reality we're in

@nina_kali_nina I'm pleased to report my static sites are legible in Dillo buuuut I might have to do something about flexbox.

@nina_kali_nina those are some great resources. ty

im glad to be mostly 'really old apps' oriented with my creativity.. Grafx2, gimp, pov etc.

gets hairy with things that filled a real niche need, like osmand/comaps -_- guess i gotta stick with older versions.

@nina_kali_nina

@nina_kali_nina *checks last release date of LCC* 24 years ago. Ok I'm still good.

@apposada I am not going to lie, I'm terrified by the changes in the FLOSS. I have nightmares, I'm in therapy, I'm taking antidepressants, and the biggest reasons of being so broken for me are the corruption of computing (which is a big part of my life) and the rise of far right (which is probably even worse, and is definitely linked to the spyware, genAI-powered killing drones etc etc). But just as much I'm broken by fear, I am also driven by it to actually try and get some control over my life. And I feel almost insane, feeling like I must be the odd one, the one who is wrong, one of the very few people who don't want any AI in their software and hardware, as long as it comes with the oppression and spyware strings attached to it the way they are attached now. I know I'm not alone, but it surely feels this way at times, especially among ex coworkers, still employed by Meta, or moving to Anthropic or OpenAI. Aaaa

@nina_kali_nina Good grief, PaleMoon and Sea Monkey! Do they still exist???

@cthos that's great; keep up the good work!

@discatte nod nod 🥰 if you'd want to write some guest blog posts about those for the blog thingy I'm hoping to start, I'll be very happy

@paulk they do, and at the very least they're getting some security patches

@nina_kali_nina I used both, so long ago... I have to revisit them.
Thank you so much, Nina.

@nina_kali_nina definitely!

@discatte @nina_kali_nina xmbdfed for bitmap fonts ;)

@nina_kali_nina@tech.lgbt is there something like browserslist that has a list of CSS/JS features supported (or not) by alternative browser projects? There are tools such as postcss and babel that can be used to turn more modern stuff like the inset shorthand back into it's older equivalent.

@mitsunee hmm, I'm not sure, I think if there isn't, it's worth making. Maybe @dillo knows?

@nina_kali_nina I tried to switch to NetBSD in July, but unfortunately the amdgpu driver only supported up to the generation before my graphics card. I learned some things from the CLI, though, and I hope they add support for my graphics hardware soon...

@nina_kali_nina You can add fcc to that although I suspect the fact the most advanced CPU it generates code for is a 6809 might make it less relevant 8)

@vaporeon_ @nina_kali_nina ANSI pcc seems to have somewhat died and while Anders popped up briefly he seems to have vanished mostly again.

@discatte @nina_kali_nina oh yay. Do you still use LCC? We used it briefly in the mid 1990s for basically an experimental build of our memory manager, on SunOS/SPARC as a sort foil against the commercial Sun CC to keep each other honest.

@nina_kali_nina @aeva Dillo is named dangerously lol

@drj @nina_kali_nina neat =^^= I mostly have experience with lcc-win, where it was one of the few ways of building a winapi app for free/easily.

@etchedpixels that's a good call-out. With all the slopocalypse I'm barely touching my 8 and 16 bit computers, and I should, as they're soul-healing.

@nina_kali_nina god, I fucking hate it here.

@mayke @nina_kali_nina can be true, but yeah us all technical people have other options, but you can't expect a journo or random other user to grep it/this, they need to up their disclamatory sections; instead of sending people into harms.

@toots @nina_kali_nina It's tragic, but most people's opinion of AI seems directly proportional to how much AI hurts them or those around them. I.e. Artists and creatives: much hate.

I'm afraid that "AI-free" is destined to be the "organic" or "free range" of software. By that, I mean it's something not easily verified, it's something where the fine print is not well understood, etc. Like regardless if you can successfully produce the product without AI, a portion of the coffee consumed during production was acquired with child labor, the PC used contains components manufactured in unsavory work conditions, the switches/routers/software that deliverer the packets were touched by AI. Yes you can say you did your best to cause no harm, but the level of pedantism necessary to truly know is unrealistic. 🥲

I don't think embracing AI is the answer either, but you need to be careful you're not acting "holier than thou" as if you go deep enough, it may be disingenuous.

@mayke @toots I think my analogy with veganism holds here. Most people aren't vegetarian or vegan, even though there are many arguments for it. From vegan point of view, harm reduction is a positive thing. Less AI is better. Simpler, more humane software is better. It might be difficult to prove things, but for small projects and communities around them, the care people put into their projects is tangible, and personal reputation is important.

I'm just very pissed off that big tech decided to convince many decision makers to put figurative meat (that also probably spies on you and rots your brain) into all the food that used to be vegan

@TheIliad have you had any luck with wsfb driver? You'd probably get a palatable desktop experience on most computers bar the most low spec ones

@nina_kali_nina RedoxOS also looks really nice, I'd be interested in contributing at some point when I get done with some of my other projects. If there was an alternative rust compiler then that would be even better cause rustc is pretty tainted

@sackofbones Rust has a relatively good anti-AI policy but fetches tainted upstream ;( and then there's LLVM.

Rust GCC is promising, but then, GCC 😓

@nina_kali_nina I cautiously recommend Chawan. It's a TUI browser that has Javascript support and is on Codeberg and Sourcehut, which are green flags in the using LLM department to me. I doubt LLMs are used in the development of Chawan, but this is a surface glance, and there's no AI policy either way.

However, it's a cautious recommendation becausr it is developed in nim, which is tainted. And iirc the nim lead (not the chawan dev iirc) is some kind of right wing asshat regardless.

@sinza that's what I've heard about Nim, yes :( I've used to be quite interested in the language a long time ago, but...

@nina_kali_nina @toots Agreed. I wish we had *one* big tech company that was confrontational and hostile against AI. Instead it's just us little people. 🤷

@etchedpixels @nina_kali_nina
I wonder if you could build cproc & QBE under Fuzix with fcc.

@kirtai @etchedpixels my NetBSD build of cproc is over a single 64k segment, so even if the language features are supported, the memory model considerations might cause problems

@will in times when I feel like this, I'd go to sdf.org, find some cool Gemini capsules or gopher phlogs, or just browse marginalia.nu catalog. 🫂

@nina_kali_nina @etchedpixels
Oh well, it was worth a thought.
(I didn't actually expect anyone to try it)

@nina_kali_nina @SRAZKVT gcc and the Linux kernel were a massive blow. As much as I swore I was done distro hopping for mission-critical machines, I feel I'm being railroaded into at least exploring BSD, especially since the distro I landed on was Debian.

To be fair, I've been BSD-curious for a while anyway. This just gives me an excuse.

@me @SRAZKVT btw it might be worth resurrecting Debian/kNetBSD. I've even recently run regular Debian 12 userland under stock NetBSD 11 kernel, and it was somewhat working through Linux emulation - I got x11 working, but no systemd. This might soften the blow for migration, maybe. Also, Debian runs very well emulated with hardware acceleration, as a way to jail it

@nina_kali_nina yeah I think slop will basically come our way, I have no involvements in the tor-browser part of anything, but I agree.

@nina_kali_nina What's wrong with gcc? I use gcc

@nina_kali_nina Also, I thought rust had one of those bullshit "responsible use" policies, or am I mistaken?

@sackofbones GCC allows LLM use in codegen if it's below 15 lines and as much as you want in tests.

Rust policy bans all LLM code: https://github.com/rust-lang/rust-forge/pull/1040/changes

@nina_kali_nina I'm not familiar with that. I pretty much wanted to at least try to play some native Linux games like Civ VI using the compatibility wrapper with my Radeon RX 6600

@wyatt @me @SRAZKVT

modload compat_linux and Carol's your auntie. There's even openSuse 15 in pkgsrc, enough to run Va-11 Hall-A (if you're lucky)

kNetBSD used to be a thing too!

@TheIliad ooh, that could be challenging :( fingers crossed your GPU will get support soon

@nina_kali_nina Oh thank fuck, so I just need to make a new llvm target. That also means I need to find a slopless C & C++ compiler, though I suppose a gcc & g++ fork from before they were tainted wouldn't be too hard to put into a state of maintenance and not of development

@sackofbones my thread mentions cproc+qbe; but I suppose old branch of llvm or gcc is still going to be required for quite some time. I remember gcc 2.95 and 3.3 used to be supported for a really really long time, I wonder if we might end up with a GCC 13 or something that will survive through the years 🥺

@nina_kali_nina And then I can have my slopless music player* if my proposal** with Dear ImGui goes through

* Cause the navidrome builtin is pretty bad and I haven't seen any others that look pretty other than Feishin, and Feishin is really buggy
** https://github.com/ocornut/imgui/issues/9558

@etchedpixels @nina_kali_nina I think the 6809 is very relevant as a survival computer in this troubled world.

@wyatt according to nia, it worked great. On my laptop with NetBSD 11 I have kernel panic triggered by libopenal 😓 see https://washbear.neocities.org/valhalla

@me @SRAZKVT

@etchedpixels@mastodon.social @nina_kali_nina@tech.lgbt I think SDCC is still fine, too? And gcc-ia16, though at least one of its downstream forks is not.

@asie @nina_kali_nina SDCC was asking about AI and what people thought. Check the dev list. I don't know what they decided as I've not really being following SDCC much of late.

@etchedpixels @asie context: https://sourceforge.net/p/sdcc/mailman/sdcc-devel/thread/3420f925-fa4c-4253-a5da-10adc9a07949%40spth.de/#msg59387747

Looks like it might end up being accepted

@etchedpixels can you point me at this "fcc" compiler? I have searched but come up empty.

@nina_kali_nina

Silver lining re compilers like Rust and GCC: their attack surface is relatively small, so once they get buried under a mountain of vibe-code tech debt, the pre-collapse versions can be forked, and civilization rebuilt from there.

But yeah, there isn't really any way to avoid compromising my ethics here. I can replace desktop environments and maybe operating systems, but I don't know of any language that can do for me what Rust can.

My only hope is to survive. Maybe.

@stylus Sorry I forgot this was going to get read by a wider circle

https://codeberg.org/EtchedPixels/Fuzix-Compiler-Kit

@f4grx I'd say 8085, 65xx and Z80 arguably will be more important due to their presence in embedded stuff. And of course avr, pic and esp, but those are likely to be out of scope for fcc?
Cc @etchedpixels

@nina_kali_nina @f4grx I have it writing code for the TMS7000 series parts so I think it can probably do most digustingly horrible chips to some level.

What it really can't cope with at this point is stuff with lots of registers although that's simply in theory a matter of writing a register allocator 8)

I think though you can make it write bad code for anything.

@nina_kali_nina personally I’m trying to think about this the same way I think about gas and animal cruelty. These systems are inherently problematic but they have been adopted to the point that ideological purity feels pretty unattainable. (It’s wild to say that about LLM’s given how new they are, but it does feel like that’s happening here too)

I hope that will change (and soon) but I (perhaps cynically) suspect it won’t. However, I do think that building knowledge, using simpler tools, refusing to engage with LLMs in a way that diminish our creativity and authorship over the things we build, all still matter. IMO they matter more than anything, because if enough people are doing that, these technologies lose so much of their destructive potential (and their market value, which is the main thing driving their destructive potential as best I can tell.)

I know I don’t have as hardline of a stance on AI as many, but it’s certainly enough to retain authorship over everything I build. I just know from experience with other societal moral failings that even when we’re stuck in an imperfect system, making it a bit better (or a bit less worse) is worth celebrating.

@slowclap I generally agree with you, but I also think that most people outside of tech can't really tell the difference between Office 2003 and Office Co-Pilot 365 and will likely happily run either, as long as it works fast enough. So I hope some of the harms could be reversed.

However, the biggest and the worst stuff is coming from big tech, either way ;( so I'm not quite sure what to do here, but I know we have to resist. I refuse to go down quietly.

@nina_kali_nina here are some notes on Debian 12.1 (aka probably the last okay mainstream linux we had) http://tilde.gay/~yottatsa/site/running-old-debian.html

@nina_kali_nina Good thread, and I'm feeling a lot of the same things - I'm even installing NetBSD today for the same reasons, after a failed attempt last night.

As for Linux, I don't think we should leap to the conclusion that low-friction *must* imply opening the doors for corporate dickweasels, even though the two often go hand in hand, and the correlation can feel inescapable. But I think the trick is, we should not let corporate interests be the ones showing up to "come to the rescue" for healing all the paper cuts. We can and should do that as a community, for our community, properly internalizing that the less technically literate are still members of our community and under our duty of care.

@MaddieM4 Good luck with NetBSD! There are quite a few good threads about installing NetBSD here on fedi, with advices and know-hows, and the documentation is generally very high quality. I hope people here can help if you run into trouble.

As for Linux, the kernel has big tech as major contributors, and it, to a large extent, serves the needs of Google, Meta, Amazon, Nvidia and so on. The voices of humans might be important to many people in the communities, but the corporate interests are going to be lobbied hard. It is already pretty bad in some parts of the kernel, I've heard

@etchedpixels oh, that's really neat! Maybe adding a new compiler target could be my fun project for one of those long winter evenings...
@f4grx

@nina_kali_nina@tech.lgbt @etchedpixels@mastodon.social

I see.

Time to go back to mourning my retro console toolchain project.

@asie I'm sorry for your (our) loss. 🫂
@etchedpixels

@nina_kali_nina@tech.lgbt @etchedpixels@mastodon.social

I mean, the goal of Wonderful Toolchain was initially to be on the frontier of whatever the homebrew/retrodev scene is putting out, and to push that frontier, right? But, every single project I have put faith in getting us there has either decided to adopt LLMs or receives no active maintenance. At the same time, I cannot justify taking a few years off my life to learn how to write an entire optimizing, GCC 6-competitive 8086 C compiler for an audience of five people, two of whom are using AI for their projects anyway. So, there's no point.

@asie @nina_kali_nina Maintenance to me is important. the 8086 gcc is amazing but it's never going to self host and about 3 people on the planet understand how it works.

For sustainable tools for old things there needs to be tooling that anyone can service.

@werdahias I haven't run properly compiled userland, just got a rootfs from Debian docker builds, unpacked it, chrooted into it with /proc mounted, and went installing things through dpkg/apt, with working around missing systemd (thus patching post install scripts on the go) and broken X forwarding (I got tigervnc and connected to it locally, like I would on Android)

@nina_kali_nina thanks for this. i might spend the rest of the year stepping through that starlight list. unless i commit to surfing the depths of geminispace 🌌
i'm glad there's still cool stuff to explore

@nina_kali_nina #DesmetC is a self hosting 8086 C compiler with no LLM code that I know of yet, and is a tiny enough project that it's unlikely to attract LLM-ers. The project owner is active and took a bunch of my patches earlier this year.

https://github.com/the-grue/OpenDC

Unfortunately, it's mostly a K&R compiler, and doesn't like most modern software. What little C89 support does exist was implemented in the early 90s in a big rush, and would benefit from being knocked down and replaced with something stabler.

@nina_kali_nina
🫂

Take it easy and do the things you want and can do. Things that give you energy and fulfill.

Be safe, be healthy, be having fun

@psf thanks for sharing! let's hope it'll stay slop-free 👍👍👍

@nina_kali_nina I didn't know Servo and it looks great, but unfortunately it's sponsored by TestMu, an AI company u_u

@kensei404 I wonder if the sponsors pivoted to AI after donating? It was ~9 months ago. Servo has a strict anti-AI policy, at the very least.

@etchedpixels @nina_kali_nina @f4grx I have postgrad experience in combinatorial optimisation, if you want to talk that over, Alan.

@nina_kali_nina @psf
It's amazing how many small, kinda hidden projects are surfacing recently.

@darkling @nina_kali_nina @f4grx That's a very kind offer although I am not sure what is possible when you see very little of the program at a time because of memory limitations or how much you need to be looking at in one go.

In theory there's enough information in the data between pass 1 and pass 2 for an optimizer of some form (or something doing other processing) to load mulltiple expression trees and the bits around them at once.

@etchedpixels @nina_kali_nina @f4grx It doesn't necessarily have to be huge on the RAM side. It might be heavy on the CPU cycles, though. There's a bunch of techniques with assorted space/time/quality tradeoffs. Pick yer poison. :)

@nina_kali_nina does zig count as a c compiler in this sense?

@sanityinc hmm, I guess it might!

@nina_kali_nina @sanityinc How about GNU Mes?

@nina_kali_nina @landley in terms of alternate OSs, how about ReactOS?

@jdboyd @nina_kali_nina I don't do windows. It's not even LP64 data model, it does an insane https://devblogs.microsoft.com/oldnewthing/20050131-00/?p=36563/ thing for legacy reasons.

Windows has been losing https://www.youtube.com/watch?v=Vhh_GeBPOhs to Linux for so long it did WSL so you could run linux binaries on windows (just like https://wiki.os2world.com/index.php?title=Win-OS/2 made sure nobody needed to write native OS/2 code back in the 90s) and is now fretting that most new "windows" software is targeted to WSL and of course https://www.windowslatest.com/2026/09/22/microsoft-keeps-rebuilding-windows-for-developers-but-a-new-poll-puts-windows-at-just-12/

Nope, zero interest.

@jdboyd sadly they're running their legacy of clean room code

@landley

@eamon @sanityinc it's probably not as fully functional as other C compilers, but, a good call, thanks!

@nina_kali_nina GCC allows for LLM-generated test cases, but nothing else according to it:
https://gcc.gnu.org/ai-policy.html

(Which is still not great especially since there are tools that can do a good job such as https://github.com/csmith-project/.)

@nina_kali_nina

Well stated. Im slowly moving to net or free for my main desktop, slowly. But yeah its not a universal solution. The 'moving to another country' metaphor is apt.

I like disruption, while discomfort in obvious ways it identifies places where ive been stuck or running unquestioned habits that would benefit from change.

And i think a lot of people think rhey want minimal, but want better automaticity, or something like that. There's a dilemma in here; to get code to that place where its stable and easy often means its also built on old paradigms, not necessarily bad or good ones, just not familiar ones.

Linux's hardware support is just flat amazing. And seamless. That's hard to beat.

@nina_kali_nina @slowclap Not just that.

I met someone doing a Masters degree who could not understand that the Windows desktop and Word were two different things.

It was just "the computer".

I think it is easy to underestimate how confusing computers can be to even otherwise intelligent people.

@etchedpixels @asie @nina_kali_nina What does sustain in this context mean? One of the advantages of a platform that old is something like the Borland toolchain is mature and nothing's going to change to break it.

@mirth @asie @nina_kali_nina That the people available at the time can understand and fix bugs realistically. So yes being so old it's got few bugs is great

@etchedpixels @nina_kali_nina @f4grx Having ruminated on it for a while (and heading for my Ballmer Peak), I suspect that there's something that works at two levels: one in the space between linear blocks (delimited by flow-control), and one at a level manipulating the blocks themselves.

Analyse the linear blocks independently and abstractly in one pass and generate in/out requirements for each, and then match those to registers/memory ops as appropriate at the higher level.

@nina_kali_nina :( all a bit grim...

@yvan there are a few more projects found in the comments, there's hope!

@nina_kali_nina "...but I still feel like I've lost my $HOME. "

That resonates here for me.

With Chat Control likely to be forced through in the EU/EEA will mean, for me at least, the death of IM and video chat.

I no longer share pictures of my offspring as I do not know who has AI agents enabled on their devices. So I do not know where those images will end up.

So the reasons why the internet seemed exciting to me in 1992, being able to discover, connect and explore seem to be disappearing.

@etchedpixels @asie @nina_kali_nina One version of this I've considered is to adopt UCSD P-System, though that has a lot of drawbacks. I want a stable platform for my journals, photos, personal records, etc, and there's a lot of gap between that and what exists already on those older platforms. That gets to be enough complexity it's hard for one person to master, but then again I also am not making my own processors so some compromise is going to be necessary.

@the_wub 😓😭🫂

@SRAZKVT @nina_kali_nina
(1) Myself I am not too worried about C compilers for flat memory addressing architectures (such as Intel x86-64).

Alas, for compilers that directly support the 8086's segment:offset addressing & far pointers, the choices become a lot more narrow. I know of
- Open Watcom
- gcc-ia16 (still has at least one codegen bug related to segmentation 🙁)
- & maybe Lindauer's cc386 (fortunately I mirrored that: https://codeberg.org/tkchia/cc386.ladsoft).

@SRAZKVT @nina_kali_nina
(2) I find "bitrot" to be a rather ... weird concept, actually.

It is like, we are blaming some Awk interpreter code (say) for not "adapting" to changes in the OS it runs on, & not asking if the OS should be making such breaking changes in the first place.

Also we really do not want to normalize people being "culturally" "expected" to maintain code in their unpaid time...

We probably need better vocabulary than "bitrot" to characterize the problem, whatever it is.

@tkchia @SRAZKVT hmm, I'm not sure if we have the same definition of "bitrot". I don't think there's a particular problem with running period correct software with period correct tooling, as long as the tooling is still available. The real problem begins when the tooling stops being available, or parts of the software needed to run stuff stop being available, right? It'd be pretty weird to expect to run idk Super Mario Bros natively on modern platforms, but if running SMB on NES required getting an un-obtanium bit of software or hardware, that'd have been a pretty bad example of bits losing their information value. I feel like I'm almost arguing for the sake of an argument, though, I haven't thought too deeply about it. Perhaps you're right.

@SRAZKVT @nina_kali_nina
(3∎) Lastly, I really should learn how to implement a device driver for USB keyboards. 🤨

@mirth hmm, it could be quite interesting, actually. Though maybe it'd be easier to implement with something like Oberon? Or a more modern Pascal system

@nina_kali_nina @SRAZKVT I suspect we actually agree. 🙂

I suppose the problem is more that bits can "rot" for many reasons. For risk evaluation purposes, it is worthwhile to reason about them separately. Un-obtainium software is definitely a real risk.

(My main beef is with the idea -- in some circles -- that a fixed piece of software "rots" & becomes less relevant over time just because "entropy", & must be constantly updated by an active (& unpaid) maintainer just to stay relevant.)

@mayke @nina_kali_nina @toots well naturally, that's how free market competition works: everyone has to fall in line with what the big guy says or else

@gardencourt @nina_kali_nina @toots We need to VHS this BETAMAX society.

@mayke @gardencourt @nina_kali_nina betamax was actually fidelity wise superiour, except their taping-lengths were supposedly pretty shorter

@toots @gardencourt @mayke @nina_kali_nina My understanding is that the thing that pushed VHS over the edge was the "extended play" or whatever mode that let you run the tape more slowly to record for a longer period of time. Without using this feature, both had similar fidelity.

I could be mistaken on this. Someone please correct me if I'm wrong.

@mayke @gardencourt @nina_kali_nina all their standard recording lengths were even below movie 90 mins to 120 mins levels but the picture fidelities were better afaik

@nina_kali_nina 🙇🖤

@nina_kali_nina I'm not sure where I found this but seems like an interesting endeavor https://www.hyperbola.info/

@nina_kali_nina @mitsunee I talked with Adële about this, it would be very useful to have a set of HTML/CSS features implemented by light browsers so you can focus on following those rules.

But building such a list is complicated. I would like to reuse many of the WPT tests so we can know which features are supported by testing them all on the browsers themselves (that's why dilloc was created).

Until we don't have a way to test these features, we don't really know how well they work.

@nina_kali_nina @werdahias I'm currently working on improving debian gnu/hurd support on debian gnu/linux. We don't use systemd in hurd and i'd be very interested in extending my cross-kernel work to also add support for debian/knetbsd

@josch you are always working on the coolest projects! 🤩

@Unlikelylass this one is the end of a huuuge yak shave -- i'm so happy the shaving is over now and i can finally actually do the work i had planned on doing a year ago :D

@nina_kali_nina I think a lot depends on priorities or constraints. For my case of an environment to preserve my personal files, the p-system approach is not really practical because I don't have the ability to write enough of the tools necessary to actually make it work (e.g. porting image codecs is just too much work). I also don't need to prioritize < 32-bit hosts. Realistically all cheap and plentiful hardware that also has a keyboard and display is 32-bit or higher.

@nina_kali_nina

as much as I enjoy "life with NetBSD" posts, save your energy for your actual needs, and please feel absolutely zero obligation to create new posts. Or even obligation to apologize for not penning a new post.

When you obtain a few spare spoons *and* decide it's worth spending them on the creation of another installment, readers can wait patiently until then.

🖖

@werdahias Better in the time until then. During hacking events I do not manage to concentrate at all and I need quiet and zero people when I work on things. During hacking events it is usually other people coming to me with the weird bugs they have with my software that I can then half-fix. XD

@nina_kali_nina What do you mean by getting started? Haven't you already started?

@ernestoDuracelli well, in a way, I have started working on it. But I want to have it working in a very specific way, and I'm not quite there yet even to say "it exists now, now I can make it better"

@lanodan
Unfortunately, I saw a post somewhere on fedi about a guy who's crawling and scraping Gemini and Gopherspace for his LLM project! We have nowhere safe! (I suppose we could limit download speed to make it harder for them, but the fact there's no tracking on the protocols actually amounts to a *bad* thing here!)
@nina_kali_nina