Flatpak is actually kinda okay for big ass /opt/ worthy gui apps as it contains it and all their dependencies in a stable environment and everything will not break and you will not waste your brandwidth with an arch linux update for example
@icedquinn@meeper bloated slow centralized rapes my home directory with its horrible layout, always mounted in the background, cripples my boot and shutdown times, wastes my ram when not in use, its ass its ass its ass
@maija@meeper flatpak has wayland issues. i could shrug but it also has gstreamer issues (mudlet is given as appimage and has no sound because of this)
@meeper@maija wayland seems like a neat idea but a bit scuffed.
at this point i'd probably say we should just make an ember profile for varlink and then use that to do some basic window management & negociating vulkan surfaces to draw to
there is also 'glow' which is a tree sync protocol on top of ember. idk if i'd want to use that part though, i haven't experimented with it enough to know if it would be good. though a desktop can be modeled as a tree so maybe?
there's also cbor but there's a few weird bits of that everyone seems to hate
@meeper@icedquinn thats just because wayland has compositing built in while on xorg its either a driver option or you need a compositor a top of it. but tearing is good sometimes, reminder that wayland only recently ever implemented it despite being unconditionally opposed to it for the longest time because games were unplayable the shitty thing about linux is that everytime it gets usable and decent and people start to switch over, shitheads like lennart poettering decide to start major components of the conventional desktop from scratch and do it even worse when they worked just fine and were never an issue the only part of this that ever was a good idea imho was pulseaudio and thats only because of the absolutely hell linux's audio shit was at the time with like a dozen competiting and conflicting standards all atop alsa
@maija@meeper you still need a protocol for sockets and shm.
XDR is basically just C structs with a well specified memory layout.
BER would be ok but it has some recursive bullshit which cbor inherited a smaller amount of. DER and Ember subset away the foot guns and isn't actually that hard to deal with.
Protobuf is kind of ok, though i prefer the way ber and cbor handle unsigned numbers
@icedquinn@meeper it seems to work, my shitty ubuntu install is stable enough using it but i still dont see why it was at all needed instead of pulseaudio which already existed
@maija@icedquinn@meeper wayland has not delivered on its simple proposition of rebuilding from the ground up learning from the past to not retain legacy features and avoid mistakes. wayland should have been a polished pearl from the beginning but instead it's missing expected functionality and creating new problems.
@meeper@icedquinn@maija ppl who made wayland came into it as the ppl with the most knowledge and best chance of learning from the history of X and they failed so there's probably no hope for the right thing
@meeper@icedquinn@maija experienced developers understand that rewrites of old software is usually a mistake and will always come with terrible regressions. sometimes it is still necessary.
@maija@icedquinn@meeper nothing beats capital and profit motive to belt out functional software fast, actual concentration on a defined goal will result in something resembling a real product. that is why you have proprietary specialized kernels from as far back as the 1980s that got work done. libre is political, we should be able to see and modify code but the libre process is slower and relies on some rando caring about a feature to get it. unless you're an entrenched libre project with corporate backing
@sun@icedquinn@meeper thats my complaint with foss in general. its just this mindset/behaiour. 99% of the devs dont know shit and at best its okay code and they just excuse horrific ux or jankiness/bugginess as "just patch it yourself" or "whats the usecase? its still usable"
@sun@meeper@maija well what they made was anemic. people kept being like we need video recording and they just went :ablobcatgooglytenor:
the BSDs wanted to support wayland but it doesn't even handle events, everyone just decided to hard dep on evdev, so now your kernel has to implement evdev for the compositor to work
@sun@maija@meeper i kind of remember there being a lot of stuff people wanted to do and they would just issue copes about it being an edge case or some future technology would provide it (like for recording the cope was gstreamer/pipewire some day after never)
but you have stuff like devilpie that still just doesn't work because they're like "ew why would somebody want a program to control windows" and its like :blobcatpain:
@icedquinn@meeper@maija nobody involved recognized that less functionality in any way will be, by users, considered a regression. they i guess considered the airtight security model to be the only goal.
@icedquinn@meeper@maija i read maling lists and i have an intuitive sense of which projects i will never participate in because the have the cancer that is spreading everywhere
@sun@meeper@maija yeah this is the same issue with flatpak. it trips over itself with too many concerns at once.
i'm also a little unimpressed with ostree. i kind of like what its trying to do, but when i looked up how to use it they talk about having to create all these intermediary branches or else it spends forever dealing with compression (and even then the compression is like really out of date and not multithreaded)
@sun@icedquinn@meeper@maija I don't think it failed technically. Just takes an awful long time to rebuild a graphic display protocol from the ground up plus adapting the existing software for it to the point where everyone already using X11 can switch without hitting a regression. And I do mean a regression, not spacebar-heating, which let's be honest tearing for games is.
And of course socially you end up with usual open-source nerds cat herding. Like OStatus→ActivityPub worked remarkably well and I think that's just because there were basically just ~3 implementations involved with few developers each and a pretty small and straightforward protocol in comparison, and most admins/users had nothing to do.
@icedquinn@maija@meeper like the balance between managing the social implications of cooperation against producing a high quality outcome is highly out of balance. every software project is necessarily a social project but it exists to produce software not to be a club
@lanodan@icedquinn@meeper@maija@sun evangelism did more harm to waymeme than its technicalities, there's a lot of shitters - even here - that were pushing it as an alternative when it was in a state at best called "unfinished" but usually where it could be more appropriately called "broken"
@sun@icedquinn@meeper 100% i like open source because security and privacy through transparency and all of that, and i dont believe in IP, but the foss model is so fucking stupid most the time imho. sure it works sometimes but, generally speaking it leads to a lot of slop. people are fucking stupid and prone to infighting bs, and without proper management and incentives they're not going to make anything good generally speaking, because why would they bother on this shit? id much rather non-foss software wih an open source be the norm, public domain license or whatnot hell proprietary if i could trust it
i'm kind of mixed because zstd is a rather large code blob. you can get close to the same performance with RoLZ with huffman tables slapped on*, and the whole ordeal is <5k SLOC. but as far as i know there isn't like a helpful deployable lib for that (there is 'orz' in rust but its kind of just done as a concept.)
i think you could do what they are doing with fastcdc (to chunk up images), a chunk store, and maybe an sccs weave to store the chunklist for revisions, though i dunno if rolz'ing up individual chunks would be useful (borg et all do this but it does kind of lose inter-chunk rendundancy stuff at the LZ layer), although a chunk store does require stash space and one of the big deployments of ostree is to shove OTA updates for embedded ewaste
@lanodan@icedquinn@maija@meeper it didnt fail forever it will eventually be good and a totally viable replacement but its just not going to be what was promised
@maija@meeper@sun yeah foss kind of demonstrates that most people are just serfs and you have a small number of people who actually organize humanity in to something respectable
unfortunately the doms also hate one another most of the time
@maija@meeper@sun the cybernetics guys kind of touch on this by saying the absolute best way to manage freedom is if everyone is also responsible for dealing with it but in practice this doesn't actually work
@lanodan@icedquinn@meeper@maija@takao every big conflict in tech that is framed as toxicity has an additional thing in common that is the real cause, unacknowledged: they were presented as an objectivly better alternative and forced into distributions way before they were functional, and people that complained were told to shut up. this is comical because up to this point a bunch of the same people were saying that floss should consider users as equal stakeholders to technical stewards, then users said hey this is unacceptable regression and they were told you are not a programmer you don't know what's best.
@icedquinn@meeper@sun agreed but dom is a weird word to use here not to get too political but i personally believe in a justified hierarchy throughout life for exactly this reason. people are just not built the same and in this case you can see the chaottic approach of foss development is just prone to infighting and ego and issues the best foss projects i've seen that werent commercial were all virtually one man jobs, ruled with an iron fist, practically managed without any foss idealism, things of this sort. sqlite being a great example of open source when you take this foss element out of it
@lanodan@icedquinn@maija@meeper@takao things like systemd had a lot of support from core and important developers and stewards of projects because I think they perceived that time for change had come but they kept replacing things that worked with reimplementations that broke shit while disregarding complaints which is a bad look.
@sun@icedquinn@meeper@maija@takao Yeah Fedora pulled that monstrous crap, really made me go like "Yup, I am even less going to ever touch this distro".
Best is like even if you'd use wayland as a daily driver, there could still be few cases where you do want an actual X11 server, specially as a developer.
@lanodan@icedquinn@meeper@maija I don't think a truly honest pitch for wayland would have flown so they necessarily had to pretend that it was just going to replace X with something betterl.
And like X11 just doesn't works well for: - Multiple GPUs, or worse hotpluggable ones - Displays with different refresh rate - Managing different colorimetry on displays (pretty important when graphic tablets got displays since) - Fractional scaling
Plus pretty sure you can't just directly encode screen content (specially something like a window) to a video stream on X11 without copying being involved, meaning higher CPU usage and much higher latency.
@sun@icedquinn@lanodan@maija@meeper I remember how amusing it was to watch Arch (btw)'s mental gymnastics in redefining what KISS and their own stupid ideological principles meant when they were pushing systemstd, good times.
you are an alternative software. you have users. there is a subset of users that are power users. they are on your offering becauese they can do something they can't do somewhere else to maximize their output. this is the sum of all the weird features your software provides. most users don't use them. you may be tempted to remove features that only a small subset of your users use. the problem is that the people that rely on these weird features are all your most vociferous defenders and advocates. they are also loudest if you remove these features. well, now youre calling your former defenders toxic and they're leaving, now whats your value proposition.
@icedquinn@lanodan@maija@meeper linux desktop is niche, your value proposition is either you are a plausible alternative to what windows or macos provides, you are probably always at a disadcantage because you're just a copy not the real thing; or, you let people customize the shit out of their desktop, which is a highly personal thing for many people. well, then you replace a major component of the desktop and a bunch of window managers and workflows etc don't work anymore. the people that are actually enthusiastic about linux on the desktop aren't enthusiastic anymore, they're angry. well, what did you expect
@sun@icedquinn@meeper@maija Which for me is horrible, because they actually are toxic on a regular basis (like freetards are well know for…), just that suddenly the targets changed.
And when it's libre software (which for me is first and foremost participatory) I find it particularly annoying because they typically won't get involved at all, so can't even ask if they are using the feature and how.
@sun@icedquinn@meeper@maija Not entirely sure it's a users thing. I think it's also that the public personalities love to be, so then the users get at least influenced into being toxic because it can seem cool or see it as something okay to be instead of reflecting on it.
@lanodan@icedquinn@meeper@maija my experience in the real world is that a lot of techy people are jjust naturally rude and hostile before they even touch a keyboard
@sun@lanodan@icedquinn@maija@meeper for many it'll be hard to become a people person after being targeted maliciously for most of their lives for various reasons