Hacker News Reader

> I mean us Linux users, trust 100% the software that we run on our machine

I don't think pretending everything is just a trusted system (in the trusted computing sense that trusted system is one whose failure would break a security policy) helps.

One of the reasons why system stability overall got better since the days of cooperative multitasking OSes is that you can no longer simply steal CPU forever or write to memory that isn't yours anymore. You will get preempted. You'll get a segfault. The OS doesn't trust, and doesn't need to trust, that all the applications running in the system are written to a quality higher that the OS itself to "just not crash all the time".

On the note of absolute trust, I have less that 100% trust on any proprietary crap that I run.

I also absolutely have less that 100% trust on any random site that I visit. I doubt anyone runs a browser that isn't under some sort of namespaces, sandboxes or MAC. So those layers of security are valuable, and I hope they continue to exist.

> To me it seems that Linux is going into a wrong direction because it makes assumptions that make sense only in the context of proprietary operating systems like macOS/Windows. We choose to ditch these systems for an open OS not to have these issues! I don't want to have the stupidity of these OS ported on Linux...

Aren't you in fact trying to shoehorn the Windows model (Just let me run this binary blob from the internet that does arbitrary side effects on my system) in Linux?

> trust 100% the software that we run on our machine

Wow! I’m sorry, this is very naive and short-sighted position, especially for a Linux user. Malware has been known to evade the open-source scrutiny and even a few days of exposure would be enough to make significant harm to certain users. Your data may not be that value or sensitive, but that’s not universally true. Most users have no idea what software is running in the Linux machines given the very complex dependency chains that exist on modern systems. Also, companies’ motivations may not be evil, but it doesn’t mean they’re in the best interest of their users, for example Canonical and their ad deal with Amazon. While I’m in favor of convenience and by no means a free software absolutist, you can’t blame the Wayland developers for trying to hold a zero-trust or trust-minimization position.

You can because it's pointless. If you run a malicious program, it can read other process memory and all the files owned by your user. The fact that it can't read another GUI window is security theater. You've made everything a lot less useful for the user without providing any security.

>it can read other process memory and all the files owned by your user

Tools like flatpak and bwrap exist. A process in a bwrap sandbox can't read other processes memory or any files that I didn't bind into the sandbox.

If I want to run an X11 GUI app in a sandbox, it can trivially escape. Wayland makes it possible to do this without a trivial escape.

Also, processes cannot read other processes memory without being a parent process or something, or having root, and with bwrap they can't even see other processes running on the system at all.

With this kind of logic, you don't fix any security hole, because you can't fix them all at once.

You can use linux namespaces or a tool like bubblewrap to sandbox the process which makes it reasonably secure. However if you're running a GUI app that rendres to an X11 server all your security gains go out the window.

> Us, and with us I mean us Linux users, trust 100% the software that we run on our machine, because we trust the open source community! Thus I don't care if any application may access the screen, because I only run trusted software.

The problem is that this mindset is incompatible with widespread adoption. The "year of linux desktop" will never come if this continues. You may be fine with this, but many other people aren't.

If the year of Linux on the desktop ever comes, we will all hate it because it'll be just like windows or macOS. Mainstream users don't want to be empowered. They want someone to hold their hand and make choices for them.

After all, look at the one Linux on the desktop option that you can buy in any store near you. It's called ChromeOS and every decision in it is made to further Google's commercial interests. It's the least open ecosystem compared to macOS and Windows (at least those can still be used without a vendor account). There is nothing there that we want. Being a niche OS is a strength, not a weakness because we Linux fans are a niche.

I agree though that 100% trusting FOSS is a fallacy. Besides the possibility of introducing malware to the repo, applications can also be exploited by malicious content like video files exploiting vulnerabilities in VLC.

Passionate Linux fans are not representative of Linux users in general.

Linux has a huge number of people who installed it on anything from grandma's old computer to a corporate server, who have exactly zero interest in any kind of tinkering with system services.

We just want something free, that won't break in an auto-update, and has access to the ecosystem of things like PipeWire and DBus.

> Mainstream users don't want to be empowered. They want someone to hold their hand and make choices for them.

I think that is deeply untrue. They want to be able to change stuff, it just turns out to be hard. I bet there's a heap of things you want to change on your car and you haven't: because you are a mainstream user.

I use Ubuntu because every time I've tried to change distro, I've been burnt.

I've got as far as finding the lines of code I need to change in "Linux", but the next steps are not simple - especially maintenance over the years. Even if one has the technical skills necessary we may choose the consumer friendly version - and it isn't because we want our hands held. I'm currently using an iPhone - nasty bugs and other terrible compromises - but for the moment the upsides are beating the downsides.

Sweeping negative stereotypes are seldomly useful.

> bet there's a heap of things you want to change on your car and you haven't: because you are a mainstream user

No. I want it to turn on and go. The less involved my car is in my life, and I with its existence, the better.

> If the year of Linux on the desktop ever comes, we will all hate it because it'll be just like windows or macOS.

This is exactly why I don't want that to happen. We're already much of the way there, though, and the trajectory is clear. That's why I'm looking into abandoning Linux in favor of BSD.

I feel the same way and it is why I moved to FreeBSD as daily driver. Very happy with it in fact.

I like it because there's much less corporate influence over it. And less need to constantly change how things work.

> Sure, from a security/privacy prospective it's good that any application can't see or interact with the content of other windows.

I got tired of hearing this so for my current desktop I went for something a bit special. I cannot log as root not "sudo" nor "su" on the desktop itself. So, from Xorg, it's not possible to have, for example, a terminal running stuff as root (so a bad actor that'd be in control of my Xorg --not something I'd like mind you-- wouldn't have an easy way to run everything he wants as root).

The way I set up is simple: I've got a laptop, without WiFi (well it has WiFi but all the drivers have been removed), on a separate LAN, that I use as my "root console". The only way to log in as root on my desktop is through SSH, from the laptop, by using a Yubikey: so I modified SSH to allow root login, but not with passwords.

That laptop doesn't have Internet access: it's only connected to the desktop.j

I also hardened the kernel running on the desktop a bit: non-root users can only see their own processes, stuff like that.

I don't know how secure it is but it was fun (and easy) to set up. I'm running that setup since eight months now: it's smooth sailing.

Besides that... I'm using X back since way before the XFree86/Xorg fork and I do actually use wmctrl and a few other tricks.

From TFA:

> The problem is that they took out many features the users depend on with no real replacement in sight.

This. I never understood the Wayland approach of telling their users, early on, "You don't want to take a screenshot of your entire screen". When I did.

I don't even know if all the Xorg features I'm using do or do not work nowadays with Wayland but I don't really care: I understood very early on that Wayland developers were actually hostile towards Xorg users.

> Xorg could have no development for another 10 years and still be more functional.

Ouch.

For me, I switched to Linux because it's what runs on the Raspberry Pi, it doesn't force auto-updates, and it's free.

For a lot of users, modularity is not a desirable feature. If you want things to work predictably every time, tinkering with stuff isn't the best plan.

I don't trust the open source community that much. Mistakes slip by all the time. For the most part, everything is fine, but people's entire lives are on computers. You get the home folder, you can probably get someone's browser password store, then you get access to someone's entire life, including not uncommonly physical access via smart locks.

Some people (not me) even have cryptocurrency that can't be recovered.

Trusting the community fully doesn't make as much sense if we want to be able to have a digital first world.

I would say that this is not really true. One of the biggest complaints with Wayland has been the lack of network transparency. Waypipe came along and added that outside the compositor. The article itself mentions pipewire adding functionality. You can still write small focussed tools.

With compositor libraries like wlroots, swc, and Louvre appearing, it is no harder to write a Wayland compositor than an X11 window manager. It looks like it will lot be long before it is easier.

And while few ever attempted making alternatives to Xorg, multiple parties have created their own Wayland compositors. You can say that nobody bothered because Xorg was great but clearly writing a Wayland compositor is far easier.

The article says multiple compositors is a problem. I see that as a benefit although I certainly get that it will be a pain for app devs. At least, it will be that way until the ecosystem matures and the bad across stand out more.

Compare the logic of what is being said here with the web. Should we all celebrate when Chrome wins the browser wars. Clients will finally have just one environment to target instead of having to deal with how multiple instances implement the standard. Oh wait, over there we think that multiple browsers is a good thing. Why not on the desktop?

No matter how shortsighted this comment is, I get the sentiment. I'll eternally miss cool x11 gimmicks that could be added to any desktop like compiz, xpenguins, xsnow and xneko

I don't think it's impossible to do that, it's just not written by anyone yet. There could be an extension to the compositor that would give your application the ability to add those gimmicks if the user chooses to allow that.