@nemesis Windows 10. I run it because I think it's fun when Windows update bricks my pc by corrupting my drivers and I have to journey across the river styx to fix it
someday if I have the right hardware I want to try running Gentoo with coreboot, kernel deblobbing, openrc, wayland, musl, clang, libressl, busybox, and doing this hardening guide, however this probably surpasses the amount of linux autism I am capable of
@mischievoustomato ehh, it fills a niche and is useful from an end user perspective but its also typical poetteringware. ideally I would like to completely avoid poetteringware and gnushit
@allison@hidden both wayland and pipewire also have the problem also that in practice using them usually requires you have some compatibility layer for x11 and pulseaudio, ideally I would likely to completely get rid of xwayland and pipewire-pulse too
I am still an nvidiacuck, mostly because my GPUs are a few years old and until fairly recently ML on not-cuda was shit and gpgpu on not-cuda is still pretty shit compared to cuda
@mischievoustomato@hidden@nemesis I very much doubt dedicated init systems are going to stick around too much longer, much less a userspace monolith like systemd. The original technical rationale for it is dead and eBPF is rapidly encroaching on many of the niches it holds.
@allison@hidden@teratology oneAPI is definitely great, it seems like something that might actually be able to compete with CUDA. it doesnt seem to me like intel has as good hardware in absolute terms (vram and tflops) as nvidia though, at least for high end stuff
@nemesis@hidden@teratology Yeah, even the top end Alchemist for consumer usage (A770 16GB, what I have) is roughly comparable in raw performance to a RTX 3060 or 3070. Whether Battlemage will change the situation or not remains to be seen.
@mischievoustomato@hidden@nemesis Unironically real and true, I've seen so many Linux true believers go through this disillusionment in real time and it's never a pretty thing to see.
we'll see. if the stuff that comes after is good and better, i dont care. At this stage, I *can't* care, since any distro i would consider worthwhile to use uses systemd, and the only one i use is tightly coupled to it.
I can imagine. At the end of the day, linux's bread and butter is the server/enterprise, first and foremost. They are who pay people to work on the thing. I think linux could be the free and decentralized thing neckbeards hoped it was, but that'd require people to change and donate money constantly to developers that already work in foss. That said, I acknowledge this isn't the case and most liekly will never be the case, and that the real change is often found in the hands of developers that are *paid* to work on foss. And as of late, I've realized that I get a lot from paid devs. Valve, RH, etc.
@mischievoustomato@hidden@nemesis Yeah for sure, I certainly don't mind paid devs doing their thing on kernel and core middleware although I do take umbrage with the GnomeOS crowd for lots of reasons. At the end of the day, it doesn't matter too much since I can always "just" jump ship to alternatives modulo hardware support (like the aforementioned Arcan and s6)
@mischievoustomato@hidden@allison honestly, being a server and enterprise thing has been great for linux imo. open source development is broken imo without traditional backing
@nemesis@mischievoustomato@hidden Especially given the ballooning complexity of hardware, non cathedral models are simply untenable unless radical simplifications are made which nobody is actually willing to do.
@nemesis@mischievoustomato@hidden As eBPF gains capabilities like instrumenting and spawning processes, I find it not outside the realm of plausibility that most of the functionality of the "traditional" PID1 will move into the kernel.
@allison@mischievoustomato@hidden huh this is a take I havent heard, can you elaborate? how could you not have a dedicated init system, something needs to be pid1
@allison@mischievoustomato@hidden ah yeah that makes sense, though the simplest way for that to happen would just be to run systemd in unikernel mode (which is possible today)
yeah. Cohesion is as important as funding. I wish there was a patreon-like kind of thing but for foss software development. I have heard a lot of bad things of the gnu stuff, but nothing I ever paid attention to beyond some performance stuff, and gpl criticisms.
@mischievoustomato@hidden@allison yup, absolutely. I dont like "open source" as opposed to free software at all, "bazaar" development doesnt work, without leadership, direction, and funding any sizeable software project is dead in the water, if it manages to make something at all it will be some bloaty disorganized hard to use mess (see: everything made by the GNU Project)
broken is an understatement. It's the backbone, the base, the support of the whole thing. Sure, hobbyists can build stuff and so on, but as stuff gets mroe complex and responsibilities pile up, it's hard to keep up
@allison@mischievoustomato@hidden the danger of supply chain attacks means some funding source will need to be created, there's way too much vulnerable mission critical dependencies and eventually corporations will want to start paying to figure out how to secure them
@allison@mischievoustomato@hidden yup, the "thousand eyeballs" idea is a myth anyways tbh, for one thing all thousand eyeballs are likely to be looking at the most interesting parts of the code and not the most vulnerable
I am almost certain the NSA has backdoors into OpenSSL and both the Linux and BSD kernels, despite them being open source
@allison@mischievoustomato@hidden well, fixing the kernels is basically impossible imo until we have simpler hardware, as you suggested
fixing userspace software *is* possible imo, it would be a large effort but I think it is doable. like producing some kind of emacs++ that you could run as PID1 and which would run everything inside an interpreter for a high level language
I think the biggest obstacle we have right now is that no HLL is good enough for the task of being the Language of Everything. I have certain ideas on what to do but, its a rabbit hole that I dont really want to go down less I fully succumb to NIH syndrome