If you need to provide tools that cross security boundaries then […] a small web app is better [than sudo].
A web app? Effin really!!? 🤨
If you need to provide tools that cross security boundaries then […] a small web app is better [than sudo].
A web app? Effin really!!? 🤨
But use the widows version and the proton layer. The Linux version is horribly coded.
I’d argue that what is holding the Linux GUI back is the amount of options, combined with the lack of proper interoperability testing (not for the lack of trying, but between the amount of options and the amount of versions, it is absolutely unfeasible), and the lack of strong design choice on the side of distributions: everyone wants to have and support everything under the sun, even if it means having 4 or 5 different flavours or editions of a particular distribution.
Don’t get me wrong, I salute the intention and the initiative, but concretely, this almost always (and I put “almost” to be safe, I’ve never seen a counter example) means a clunky, unpolished experience in most cases.
I usually describe it as:
If GUIs were doors:
- Mac OS would be selling literally only one kind of door, that is super slick, brushed metal, glass and white, fancy, with a black glass and brushed metal handle, has a great feel to it, good heft, great handling, satisfying sound and feedback, etc, but then you need to buy everything else from them (including your lights, flooring, etc) or it just won’t open. Of course they sell everything at a premium.
- Windows would be your standard wooden office door with the standard metal handle and the standard automatic door closer; but anyone can open it even when locked, it needs to be changed every other year, if you “customise” (i.e. adapt it in any way) it it will wear out 10x faster, and any adjustment you do (handle spring tension, automated closer strength and kickback, hinges adjustment, etc) will be reset at night randomly every other week, the door will get new “features” (like microphones, a search prompt, an assistant, etc) randomly, and you can use any kind of furniture you want, but during the “night resets” (aka “upgrades”), all the furniture in the office will be reset to be “Microsoft furniture”, and you will need to exchange it all back in the next morning. And for various unpredictable reasons, once in a while, when going through the door, it will close unexpectedly and violently, slamming you in the face with full force.
- Linux and FOSS in general is a collection of community made IKEA inspired doors. You can mix and match anything. Any kind of door, any kind of hinge. Any kind of handle. Want a door that opens sideways? Go for it. Want a door that slides up? Do it. Want a butterfly door? Sure. A proximity sensor as a handle? Totally. A carbon fibres and ceramic door? Absolutely. All at once? Why not. In the end, no door is exactly the same, even across the same building, and you often need a few minutes to figure out how new doors work in new buildings. And of course, lots of doors are ill designed, with completely unnecessary features, and conflicting options, like both a sideways and butterfly hinge. Still works, but has caveats. But hey, if it breaks, or doesn’t fit, you can change it any time, get parts anywhere, and there is an absolutely insane amount of community made documentation on most of it (except the internals, some of it is hard to understand, some of it is absolutely obscure, and most of it is documented by people who made it exclusively for people who made it)
IMHO what we would need is for distributions to “adopt” a given GUI (or DE), and stick to that. Do not even carry the packages for something else. If it is needed, another distribution will be made. That would simplify things a lot, and would greatly relieve the stress on maintainers.
And it would make for a much more approachable user experience.
And Docker initially used Ubuntu. They explicitly and specifically switched to Alpine in 2016 for performance, to minimise the overhead.
Note: this comment is long, because it is important and the idea that “systemd is always better, no matter the situation” is absolutely dangerous for the entire FOSS ecosystem: both diversity and rationality are essential.
Systemd can get more efficient than running hundreds of poorly integrated scripts
In theory yes. In practice, systemd is a huge monolithic single-point-of-failure system, with several bottlenecks and reinventing-the-wheel galore. And openrc is a far cry from “hundreds of poorly integrated scripts”.
I think it is crucial we stop having dogmatic “arguments” with argumentum ad populum or arguments of authority, or we will end up recreating a Microsoft-like environment in free software.
Let’s stop trying to shoehorn popular solutions into ill suited use cases, just because they are used elsewhere with different limitations.
Systemd might make sense for most people on desktop targets (CPUs with several cores, and several GB of RAM), because convenience and comfort (which systemd excels at, let’s be honest) but as we approach “embedded” targets, simpler and smaller is always better.
And no matter how much optimisation you cram into the bigger software, it will just not perform like the simpler software, especially with limited resources.
Now, I take OpenRC as an example here, because it is AFAIR the default in devuan, but it also supports runit, sinit, s6 and shepherd.
And using s6, you just can’t say “systemd is flat out better in all cases”, that would be simply stupid.
Larger might be acceptable too, not sure
It should.
Also, work off of the copy. Never touch the source.
This isn’t about you. Your data is key in making other, relevant people stand out. Why relevant? Because they resist. Because they aren’t depressed, they are fighting back. And whatever the reason that drives them, they can be manipulated. By compromising their anonymity. By making literally everyone else a “known variable”. With your help.
Better lzma performance with xz. 🤪
Have fun with the initramfs.
“I’m afraid that’s not a choice…”