Conversation

Notices

1. Embed this notice
   Wolf480pl (wolf480pl@mstdn.io)'s status on Sunday, 22-Oct-2023 17:57:37 JST Wolf480pl

   • Haelwenn /элвэн/ :triskell:
   • iced depresso

   @novenary @icedquinn @lanodan
   I want to be mad, but honestly, with Linux kernel's "no regressions"approach, I'm fine with 2 years.

   Still, I don't like the trend and the message it sends. Other projects that are not backwards compatible may take it as a hint that they don't need 5 year LTSes. And I don't want to do backwards incompatible updates of everything every 1-2 years

   • Embed this notice
     iced depresso (icedquinn@blob.cat)'s status on Sunday, 22-Oct-2023 17:57:36 JST iced depresso

     in reply to

     • Haelwenn /элвэн/ :triskell:
     @wolf480pl @novenary @lanodan i think lts stability is probably necessary, though i couldn't tell you how long of a window is good.
     
     mostly from my own experience dealing with indie and commercial devs. the linux platform is too unstable to work with unless you're going to have people who's job is to just literally sit there and run regression tests all the time.
   • Embed this notice
     iced depresso (icedquinn@blob.cat)'s status on Sunday, 22-Oct-2023 18:08:21 JST iced depresso

     in reply to

     • Haelwenn /элвэн/ :triskell:
     @wolf480pl @novenary @lanodan desktop software, games.
     
     most of the commercial software was shipped for cent/ubuntu. indeed, a lot of research stuff too because those guys have better things to do than distro politics and workarounds.
   • Embed this notice
     Wolf480pl (wolf480pl@mstdn.io)'s status on Sunday, 22-Oct-2023 18:08:22 JST Wolf480pl

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • iced depresso

     @icedquinn @novenary @lanodan
     what kinda devs / usecase you mean? Like making an appliance that ships with Linux? Or more like making server or desktop applications?

   • Embed this notice
     Wolf480pl (wolf480pl@mstdn.io)'s status on Sunday, 22-Oct-2023 18:15:10 JST Wolf480pl

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • iced depresso

     @icedquinn @novenary @lanodan
     Oh so you mean the whole OS is not a stable platform, not the kernel itself.

     Yeah, definitely.
     Most ofbthe userspace doesn't have as strong a stance on backwards compat as the kernel, enen though they IMO should.

   • Embed this notice
     iced depresso (icedquinn@blob.cat)'s status on Sunday, 22-Oct-2023 18:15:10 JST iced depresso

     in reply to

     • Haelwenn /элвэн/ :triskell:
     @wolf480pl @novenary @lanodan yeah we may need something like an immutable base system and then some kind of nix/guix thing on top of that, though that's still bodging around the problem pretty hard.
     
     flatpak tries to address the same issue--but runs in to a big stupid in that without an LTS there is no stability there either. every program requires its own gnome install :ablobcateyeroll:
     
     well, elementaryos fixes that. they use flatpak with their own base images. so they are forcibly stabilizing it, but they do that for themselves.
   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Monday, 23-Oct-2023 01:53:07 JST Haelwenn /элвэн/ :triskell:

     in reply to

     • iced depresso
     @icedquinn @novenary @wolf480pl Libre software should link themselves to a distro, the packaging infrastructure is already there to run tests and stabilize things out accordingly.
     And you could maybe also imagine something similar to CPAN Testers.
     Which is the right way of doing things, the "ship a thing once, never touch it again" only works on things like releases dedicated to a video game console where both the hardware and the software barely ever changes.
     
     (And I don't give a shit about proprietary software, I consider that to be broken by default and stuck broken)
   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Monday, 23-Oct-2023 03:08:31 JST Haelwenn /элвэн/ :triskell:

     in reply to

     • iced depresso
     @wolf480pl @novenary @icedquinn I was sure it was going to lead to "but this distro sucks" moment.
     
     Get better at choosing distros, don't listen to people like tech journalists that can't look past the fucking desktop if you're going to choose something you'll use for your business or most of your days.
     Priorities for choosing a distro ought to be something like:
     1. Do you like the community
     2. Do you like the package manager and related specific tools
     3. Do you like most of the various policies
     4. Do you like the defaults of lack thereof
   • Embed this notice
     Wolf480pl (wolf480pl@mstdn.io)'s status on Monday, 23-Oct-2023 03:08:32 JST Wolf480pl

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • iced depresso

     @lanodan @novenary @icedquinn
     and then you run pacman -Syu and it's like
     "libwhatever was replaced by libwhatever2, wanna install it instead? [Y/n]"
     "foo depends on libwhatever, wanna uninstall foo? [y/N]"

   • Embed this notice
     iced depresso (icedquinn@blob.cat)'s status on Monday, 23-Oct-2023 03:10:20 JST iced depresso

     in reply to

     • Haelwenn /элвэн/ :triskell:
     @wolf480pl @lanodan @novenary i wonder what the performance overhead of CLAP's api factories are. or if that's something that could be fixed at the linker/loader level if it is indeed an issue.
     
     they have an interesting approach where the first call you make is to ask for an entry point by string name, so its like
     
     interface = entry("quinndows/3.11")
     
     and you either get null or the interface (struct full 'o pointers, but i think in clap this just gives you the plugin factory where you then ask for modules out of the library)
     
     though that may not be sufficiently different from how we link to libraries as it is, considering we're supposed to put symlinks to compatible .so version. idk.
   • Embed this notice
     Wolf480pl (wolf480pl@mstdn.io)'s status on Monday, 23-Oct-2023 03:10:21 JST Wolf480pl

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • iced depresso

     @lanodan @novenary @icedquinn
     IOW, at some point the breaking API changes reach your app, and if you're not there to update to the newer versions of the libs (inevitably introducing new bugs), you're fucked, and the distros won't help you

   • Embed this notice
     Wolf480pl (wolf480pl@mstdn.io)'s status on Monday, 23-Oct-2023 03:12:00 JST Wolf480pl

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • iced depresso

     @icedquinn @novenary @lanodan
     An immutable base system is a workaround, and a costly one: consider all the copies of webp parser that will never be updated and will stay vulnerable forever.

     A proper solutions would be ABI compatibility guarantee from userspace libraries, like the one you get from Linux kernel

   • Embed this notice
     iced depresso (icedquinn@blob.cat)'s status on Monday, 23-Oct-2023 03:12:00 JST iced depresso

     in reply to

     • Haelwenn /элвэн/ :triskell:
     @wolf480pl @novenary @lanodan i wrote an actual book on that https://iceworks.cc/g/immortal-abis/ but pretty much nobody cared.
     
     trying to point out how everyone who isn't linux offers platform stability results in retarded bullshit from the linux community.
   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Monday, 23-Oct-2023 03:13:22 JST Haelwenn /элвэн/ :triskell:

     in reply to

     • iced depresso
     @wolf480pl @novenary @icedquinn They absolutely can.
     
     In many distros if a library breaks applications, either said library doesn't gets stabilized, or it gets patched.
     And when a new major API gets released the packagers can lend a hand to help, specially when it comes to testing things out.
     (That said when choosing dependencies you should know about them, because they're a massive liability)
     
     This is why distro devs typically have commits in a whole bunch of upstreams.
   • Embed this notice
     iced depresso (icedquinn@blob.cat)'s status on Monday, 23-Oct-2023 03:13:30 JST iced depresso

     in reply to

     • Haelwenn /элвэн/ :triskell:
     @wolf480pl @novenary @lanodan bag of pointers is how you do interfaces in C :blobcatdunno: the civilized world... just has actual interface abstractions.
   • Embed this notice
     Wolf480pl (wolf480pl@mstdn.io)'s status on Monday, 23-Oct-2023 03:13:36 JST Wolf480pl

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • iced depresso

     @icedquinn @novenary @lanodan
     > struct full of pointers

     please don't

   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Monday, 23-Oct-2023 03:15:50 JST Haelwenn /элвэн/ :triskell:

     in reply to

     • iced depresso
     @wolf480pl @novenary @icedquinn Stuff *is* going to break, bugs happens, always.
     What matters if how easy/fast they get fixed and getting the proper infrastructure to prevent them from reaching users.
     
     And if you go anti-distro you're going to keep long time bugs (as seen by your point with libwebp).
   • Embed this notice
     Wolf480pl (wolf480pl@mstdn.io)'s status on Monday, 23-Oct-2023 03:15:51 JST Wolf480pl

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • iced depresso

     @lanodan @novenary @icedquinn
     I've had Debian stable push a backwards-incompatible fix for unbound DNS server as a *security update*. It broke my stuff.

     If devs don't keep backwards compatibility, there's only so much a distro can do.

   • Embed this notice
     iced depresso (icedquinn@blob.cat)'s status on Monday, 23-Oct-2023 03:22:31 JST iced depresso

     in reply to

     • Haelwenn /элвэн/ :triskell:
     @wolf480pl @lanodan @novenary i don't mind people using rolling stuff, i just have direct experience in how LTS releases are valuable.
     
     failing those we are stuck with having to work out copes with nix/guix and both of those options take over your entire system :\
   • Embed this notice
     Wolf480pl (wolf480pl@mstdn.io)'s status on Monday, 23-Oct-2023 03:22:33 JST Wolf480pl

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • iced depresso

     @lanodan @novenary @icedquinn
     I'm not anti-distro, what I'm saying is you also need upstream devs to care about backwards compat.

     > how easy/fast they get fixed

     what if the answer is "sorry we couldn't fix this one"?

     And like, I get it, distros are doing their best.

     But if RedHat and Suse are giving up, it seems like it's getting harder and harder for them.

     On top of that you get shittons of libraries bypassing distros, from pip & npm to rust & golang.
     1/

   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Monday, 23-Oct-2023 03:28:06 JST Haelwenn /элвэн/ :triskell:

     in reply to

     • iced depresso
     @wolf480pl @novenary @icedquinn language-specific package managers cannot ever work right, which is why I avoid them like the plague.
     
     Not to mention that pip cannot even resolve dependencies correctly and if you have any care about security, you shouldn't be using it (setup.py means arbitrary code being executed, even if you simply want to download without installing; pypi contains a whole bunch of binaries, good luck reviewing that).
     
     The simple reason for this is because they have 0 integration with the rest of the system, for example if a library bumps ABI without breaking API, the overlay package manager has no information about it and cannot do things like automatically rebuilding. (Something you get on Gentoo for example)
   • Embed this notice
     Wolf480pl (wolf480pl@mstdn.io)'s status on Monday, 23-Oct-2023 03:28:51 JST Wolf480pl

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • iced depresso

     @lanodan @novenary @icedquinn
     The landscape is getting worse and AFAIU it's because of the developers' attitude.

     Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     iced depresso (icedquinn@blob.cat)'s status on Monday, 23-Oct-2023 03:30:27 JST iced depresso

     in reply to

     • Haelwenn /элвэн/ :triskell:
     @lanodan @novenary @wolf480pl outside of microsoft and to a lesser degree apple, there is basically zero traction for ABI stability.
     
     if i recall rust never even bothered to specify one.
     :blobancap: :blobcattrans: :blobancap: :blobcattrans: :blobancap: :blobcattrans: likes this.
   • Embed this notice
     iced depresso (icedquinn@blob.cat)'s status on Monday, 23-Oct-2023 03:33:32 JST iced depresso

     in reply to

     • Haelwenn /элвэн/ :triskell:
     @wolf480pl @novenary @lanodan :blobcatdunno:
     
     i know nim does but it.. gets a little fucky. the object pascal lineage has never taken well to shared objects and while it has been made to do so at times it tends to have weird caveats like needing to link with the same runtime as the other modules, so in practice it only does for sharing modules with the same assembly :neocat_woozy:
   • Embed this notice
     Wolf480pl (wolf480pl@mstdn.io)'s status on Monday, 23-Oct-2023 03:33:33 JST Wolf480pl

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • iced depresso

     @icedquinn @novenary @lanodan
     does Rust even have dynamic linking?

   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Monday, 23-Oct-2023 03:39:37 JST Haelwenn /элвэн/ :triskell:

     in reply to

     • iced depresso
     @wolf480pl @novenary @icedquinn I would never touch a 5 year LTS distro other than maybe as part of a dayjob where I would need to be careless as fuck to not want to do things right.
     
     The BSD way of freezing the core system (where they actually can and do ABI stability) and then having the ports being more or less rolling depending on your flavor make a lot more sense.
     
     Because you can't freeze everything, specially given that some dependencies need regular updates.
     
     You need to embrace at least a part of what people are doing if you're going to use their software.
   • Embed this notice
     Wolf480pl (wolf480pl@mstdn.io)'s status on Monday, 23-Oct-2023 03:39:38 JST Wolf480pl

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • iced depresso

     @lanodan @novenary @icedquinn
     Yes, language package managers are horrible.

     The problem is, upstreams use them.
     Upstreams make useful software that sits on heaps of dependencies that are not packaged by any distro.
     Dependencies that break backwards compat twice a year, and provide no bugfixes for prior versions.

     Try packaging that for a 5 year LTS distro.

   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Monday, 23-Oct-2023 03:44:08 JST Haelwenn /элвэн/ :triskell:

     in reply to

     • iced depresso
     • Johnny Peligro
     @mischievoustomato @novenary @icedquinn @wolf480pl It's really not. Said blogpost is basically gaslighting similarly to a fucking advertisement. Easy Anti Cheat was depending on what's pretty much spacebar heating and someone that's being honest would also acknowledge that Windows/Wine/… breaks those kind of hack regularly, because you cannot ever not have regressions on thing sneakily depending on side effects.
   • Embed this notice
     Johnny Peligro (mischievoustomato@5dollah.click)'s status on Monday, 23-Oct-2023 03:44:09 JST Johnny Peligro

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • iced depresso
     @icedquinn @novenary @wolf480pl @lanodan tfw the stable base for linux is... win32
   • Embed this notice
     iced depresso (icedquinn@blob.cat)'s status on Monday, 23-Oct-2023 03:46:36 JST iced depresso

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • Johnny Peligro
     @lanodan @mischievoustomato @novenary @wolf480pl dt_hash is part of the elf standard. glibc is using some yee-haw shit.
   • Embed this notice
     Wolf480pl (wolf480pl@mstdn.io)'s status on Monday, 23-Oct-2023 03:54:34 JST Wolf480pl

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • iced depresso
     • Johnny Peligro

     @mischievoustomato @novenary @icedquinn @lanodan
     yeah, the driver end of the kernel doesn't quite fulfill that promise...

     but at least the userspace ABI doesn't break, so if the kernel works fine with your hardware, you should be able to use as old a libc on it as you want...

     except for libgl I guess

   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Monday, 23-Oct-2023 03:54:34 JST Haelwenn /элвэн/ :triskell:

     in reply to

     • iced depresso
     • Johnny Peligro
     @wolf480pl @mischievoustomato @novenary @icedquinn Yeah, there's two ways:
     - Containers: The kernel is somewhat stable (in reality you're likely going to hit weird issues). While userland, including the libc just isn't. Problem is, you've become a distro and now you ought to care about *everything* in your tree (like updating log4j or libwebp);
     
     - Packages: You declare your dependencies, write a proper testsuite to catch bugs early, care about what downstream tells you (like don't be calibre and claim that python2 is fine). Of course that one can be a bit annoying depending on the particular distro, which is why they ought to improve;
   • Embed this notice
     Johnny Peligro (mischievoustomato@5dollah.click)'s status on Monday, 23-Oct-2023 03:54:35 JST Johnny Peligro

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • iced depresso
     @wolf480pl @novenary @icedquinn @lanodan ah yeah
     i found that statement funny because a lot of kernel updates broke stuff for me in terms of hardware functionality 👊😎
   • Embed this notice
     Wolf480pl (wolf480pl@mstdn.io)'s status on Monday, 23-Oct-2023 03:54:36 JST Wolf480pl

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • iced depresso
     • Johnny Peligro

     @mischievoustomato @novenary @icedquinn @lanodan

     Linus Torvalds's official stance is that if a kernel update breaks userspace, that's a kernel bug.

   • Embed this notice
     Johnny Peligro (mischievoustomato@5dollah.click)'s status on Monday, 23-Oct-2023 03:54:37 JST Johnny Peligro

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • iced depresso
     @wolf480pl @novenary @icedquinn @lanodan >no regressions approach
     what
   • Embed this notice
     iced depresso (icedquinn@blob.cat)'s status on Monday, 23-Oct-2023 03:55:18 JST iced depresso

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • Johnny Peligro
     @mischievoustomato @novenary @wolf480pl @lanodan the big win with nix is that it can cope with incompatible histories.
     
     like the time i tried to use software that required old versions of dependencies that couldn't coincide on the same system and had to be exiled off to /opt if they could be used at all
   • Embed this notice
     Johnny Peligro (mischievoustomato@5dollah.click)'s status on Monday, 23-Oct-2023 03:55:19 JST Johnny Peligro

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • iced depresso
     @wolf480pl @novenary @icedquinn @lanodan back when I didn't run nix i would just cope and seethe and live with the regression or run an older kernel/mesa
     now i run nixos and can either revert/boot to an older generation or easily patch out regressions
   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Monday, 23-Oct-2023 03:58:01 JST Haelwenn /элвэн/ :triskell:

     in reply to

     • iced depresso
     • Johnny Peligro
     @icedquinn @mischievoustomato @novenary @wolf480pl Even if GNU would have sent proper documentation to the ones responsible for the ELF standard (USL) and getting it accepted, it would probably still have happened.
     
     Also the one hitting the regression was IIRC only Arch, which is basically just a testground distro where you hit regressions on a frequent basis.
   • Embed this notice
     iced depresso (icedquinn@blob.cat)'s status on Monday, 23-Oct-2023 04:01:50 JST iced depresso

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • Johnny Peligro
     @lanodan @mischievoustomato @novenary @wolf480pl it hit gentoo and void as well. ended up switching to flatpak version of steam (where they do basically forcibly stabilize an LTS distro on their own.)
     
     i refuse to blame people who implement a spec when other people yee-haw random shit and act like this is acceptable behavior (i also sent displeased glares at the coreclr repo for doing the same; they basically said yeah we aren't going to update ECMA-335 because fuck you)
   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Monday, 23-Oct-2023 04:16:47 JST Haelwenn /элвэн/ :triskell:

     in reply to

     • iced depresso
     • Johnny Peligro
     @icedquinn @mischievoustomato @novenary @wolf480pl
     Meanwhile I take it like this:
     > Thing about removing something legacy
     > Works fine for everything we can run tests on / can scan source code
     > Do it
     > Some proprietary shit breaks
     
     Gamers™ reaction: Noooo linux isn't stable! Support the things better!
     
     My reaction all the time: Stop using proprietary shit, there cannot ever be proper community support for those, they literally forbid it by both law and not sharing the means of support.
     
     And honestly if you think standards/specs specially the ones managed by some random committee/organisation are something actually followed you're going to be constantly displeased because working with standard bodies at best a pain, at worst pretty much impossible (like either because they're stuck in time / gatekeepers, or because they're pipe dreamers who can't grasp implementers needs).
   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Monday, 23-Oct-2023 04:20:41 JST Haelwenn /элвэн/ :triskell:

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • iced depresso
     • Johnny Peligro
     @icedquinn @mischievoustomato @novenary @wolf480pl Also it's a funny thing because it likely means there was a massive blind spot for EAC.
     AFAIK the ld.so of glibc will load gnu_hash when present and so ignore dt_hash.
   • Embed this notice
     iced depresso (icedquinn@blob.cat)'s status on Monday, 23-Oct-2023 04:31:33 JST iced depresso

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • Johnny Peligro
     @lanodan @mischievoustomato @novenary @wolf480pl well they didn't publish their yee-haw'd shit in their own technical report either, so.
   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Monday, 23-Oct-2023 04:38:49 JST Haelwenn /элвэн/ :triskell:

     in reply to

     • iced depresso
     • Johnny Peligro
     @icedquinn @mischievoustomato @novenary @wolf480pl What do you mean exactly by technical report?
     
     Because it's been documented for at least 10 years by now (just not by GNU AFAIK) and with third-party implementations like LLVM's lld, musl's ld.so, …
     It's very much a de-facto standard.
   • Embed this notice
     iced depresso (icedquinn@blob.cat)'s status on Monday, 23-Oct-2023 04:40:04 JST iced depresso

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • Johnny Peligro
     @lanodan @mischievoustomato @novenary @wolf480pl
     > just not by GNU
     and there is no excuse for that.
     
     "just read the browser source code to learn how javascript works bro" :neocat_gun: not acceptable.
   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Monday, 23-Oct-2023 04:46:30 JST Haelwenn /элвэн/ :triskell:

     in reply to

     • iced depresso
     • Johnny Peligro
     @icedquinn @mischievoustomato @novenary @wolf480pl No source code reading needed though, just non-official sources.
     And given the context of EAC, they likely need to reverse-engineer code on a regular basis to figure things out.
   • Embed this notice
     iced depresso (icedquinn@blob.cat)'s status on Monday, 23-Oct-2023 04:48:43 JST iced depresso

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • Johnny Peligro
     @lanodan @mischievoustomato @novenary @wolf480pl i don't really care about the context of whatever.
     
     they've had 14 years to document their shit. they haven't done it.
     
     this is just plainly unacceptable behavior for core systems.
   • Embed this notice
     narcolepsy and alcoholism :flag: (hj@shigusegubu.club)'s status on Monday, 23-Oct-2023 04:50:37 JST narcolepsy and alcoholism :flag:

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • iced depresso
     • Johnny Peligro
     @lanodan @mischievoustomato @novenary @icedquinn @wolf480pl idk man, long-term support for legacy stuff is overall win for everybody IMO.
   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Monday, 23-Oct-2023 04:52:17 JST Haelwenn /элвэн/ :triskell:

     in reply to

     • iced depresso
     • Johnny Peligro
     @icedquinn @mischievoustomato @novenary @wolf480pl GNU basically never documents their extensions, this is why I say they do vendor-locking.
     But that should only matter for ones where you don't have proper third-party implementations/documentations, which here you absolutely did.
     
     Like if EAC wanted, they could be like "Nope, we're doing ld.so ourselves" (probably a terribly bad idea btw) and they would be able to implement gnu_hash support easily.
   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Monday, 23-Oct-2023 04:54:45 JST Haelwenn /элвэн/ :triskell:

     in reply to

     • iced depresso
     • narcolepsy and alcoholism :flag:
     • Johnny Peligro
     @hj @mischievoustomato @novenary @icedquinn @wolf480pl You'd want PleromaFE to support 5 years old Firefox/Chrome? I'd love to see you try.
     
     It's a loose-loose situation, you end up with pile of hard-to-test workarounds (specially when proprietary systems are involved) on both sides.
   • Embed this notice
     narcolepsy and alcoholism :flag: (hj@shigusegubu.club)'s status on Monday, 23-Oct-2023 05:05:18 JST narcolepsy and alcoholism :flag:

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • iced depresso
     • Johnny Peligro
     @lanodan @mischievoustomato @novenary @icedquinn @wolf480pl >You'd want PleromaFE to support 5 years old Firefox/Chrome? I'd love to see you try.
     
     I'm pretty sure it does support 5yo Firefox/Chrome. There's also always graceful degradation.
     
     Graceful degradation over complete breakage.
   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Monday, 23-Oct-2023 05:17:31 JST Haelwenn /элвэн/ :triskell:

     in reply to

     • iced depresso
     • narcolepsy and alcoholism :flag:
     • Johnny Peligro
     @hj @mischievoustomato @novenary @icedquinn @wolf480pl Try it, I'm pretty sure it doesn't works and actually supporting it would be horribly annoying.
     
     Specially because graceful degradation isn't much of a thing in JavaScript, which is why horrors like babel and core-js exists, but stuff still falls through the cracks once in a while.
     
     And that's for the web, where things like ABI issues simply do not exist (except maybe with WASM) and you can quite nicely query for an API existence, which pretty much doesn't exists in other ecosystems.
   • Embed this notice
     iced depresso (icedquinn@blob.cat)'s status on Monday, 23-Oct-2023 05:36:27 JST iced depresso

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • narcolepsy and alcoholism :flag:
     • Johnny Peligro
     @lanodan @mischievoustomato @novenary @wolf480pl @hj .NET has some provisos for that (you can specify library versions in an assembly's dependency header), not sure what the JVM does, and you can do degredation in C but there is a caveat: you have to pull your stuff in with dlopen/dlsym or use compiler-specific shenanigans like the weak linker flag (JACK uses this.)
     
     CLAP and LV2 also covered quite a bit of that, one by using just C constructs to get factories and the other using RDF graph manifests. LV2, kinda, ends up replicating .NET's metadata model but using rdf turtles instead of a bespoke binary format.
     
     so.. it's all doable, its just that people who get paid to do it work on proprietary software where all the money is :comfywoozy:
   • Embed this notice
     narcolepsy and alcoholism :flag: (hj@shigusegubu.club)'s status on Monday, 23-Oct-2023 06:08:29 JST narcolepsy and alcoholism :flag:

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • iced depresso
     • Johnny Peligro
     @lanodan @mischievoustomato @novenary @icedquinn @wolf480pl >actually supporting it would be horribly annoying.
     
     Tell me about it. I've done webdev for IE6 and Undebuggable jank android shit.