Oh! Came up with a new one, though it’s more of a unixism than a Linux specific thing.

I really wish that the core utils and other cli tools had a standard structured output option, like yaml, json, or toml so that it would be easier to parse rather than all of the random regular expressions needed when piping output around.

Edit: And it would be great if we also picked that same format for config files instead of all the bespoke stuff in /etc.

Its users

• audio - Most of the time it works, but there have been plenty of times that after an install, I have to go in and make a handfull of changes to get it working.

• “you are using it wrong” developers - Lookin at you, Gnome, Mozilla and Pottering. Yes, you are donating your time, and I appreciate that, but don’t be dismissive of people if they bring up valid issues. If you just don’t want to fix problems, that’s fine, but just be honest about that, instead of blaming the user.

• sleep/hibernate - I’ve never depended on sleep or hibernate to work properly. I gave up on that years ago, and whenever I come back and try it again, I remember why I gave it up.

• documentation - As a seasoned linux person, I love man-pages, but they are soooooo obtuse and hard to parse for newbies. I also hate it when the website has mountains of documentation, but they couldn’t be bothered to put that into the man-pages.

• video/wifi drivers - Yes, I know that this is mostly a problem because of the manufacturers. That doesn’t mean it isn’t a problem.

• unsympathetic users - Just because it works for you, doesn’t mean it works for other people. I can’t wait for year-of-the-linux-desktop, but it just isn’t there yet. As soon as you have to tell a non-tech to open a terminal, the vast majority of them are out. You and I know that ‘editing /etc/somedir/somefile and running /usr/sbin/somecommand’ is easy, but sooooo many of them don’t know what that means, nor will they care. I hear that windows is pretty bad nowadays, but people will often stick with the devil they know.

We have awesome distributed systems like Kubernetes (rke2, or k3s as easy distro examples) BUT no desktop usage.

I want a distributed desktop dang it. My phone, my smart tv (media PC), my gaming computer, my SOs gaming computer, my router, my home lab, etc, etc should theoretically all be one computer with multiple users, and multiple interfaces.

I have spent way too much time fiddling with audio, both in PulseAudio and Pirewire. Granted, this sucks even more on Windows.

Weird how my absolute favorite thing about Linux is how easy and simple installing software is, at least on Arch. Never touched a flatpack or snap or whatever else they’re called for my 13+ years if use.

The community’s general overestimation of the average person’s tech capabilities.

Not necessarily fair to pin this on Linux per se, but there’s hardware that doesn’t work well or at all still and alternative solutions still aren’t there. So this would be mostly on companies making software for Windows but not for Linux, but it’s still part of the Linux experience that I do not enjoy.

I have to troubleshoot things on Linux more than I did on Windows.

My one major complaint is audio in general. I’ve had so many audio issues. If you need an eq or noise canceling it’s a pain to get it working. There’s always a bug somewhere, always a random distortion.

Voicemeeter is the only thing I miss about Windows. I really do.

Having to install apps manually and figure out dependencies myself because a popular piece of software only officially supports Ubuntu and Debian. No normal human would ever do this. They would go back to Windows. Hell, I still haven’t even gotten one piece of software to work on my new OpenSUSE system yet: Beyond Compare 4. There’s no flatpack for it. The RPM test says all dependencies are satisfied, but when I run the program, nothing happens. I did some web searching, but I haven’t dug too deep yet.

Why are there so many package managers with such different syntaxes? And why does one repo maintainer decide to call it “package” and another calls it “package4”? Or some entirely different name! It’s maddening. I’ve had to create empty proxy packages that translate package names just to install some RPM file. Again, the average person is not going to do this.

In KDE plasma, the first thing most people do is set up Wi-Fi on their computer, but you need to set up KWallet first or else the password gets stored in some other dimension. I accidentally typed my Wi-Fi password wrong, and I couldn’t for the life of me figure out how to clear it out and make it ask me for the proper password when I try to connect. I even went into network manager and switched the network to say, “ask me every time”. It wouldn’t! It would just sit there and hang on “authenticating”. I never did figure it out. I ended up forgetting to encrypt my system partition, so I simply reinstalled the OS.

Terrible documentation that is written assuming far too much prior knowledge.

I’m pretty technologically literate but just don’t have a lot of experience with Linux, in the last year of trying properly to switch over the most frustrating part is trying to fix problems or follow peoples “guides” to various things. There is plenty of information out there for sure but when I have to keep looking up a string of things to try and get to my desired end result then the original documentation I’m trying to follow is not adequate.

I can only imagine what it might be like for users who are less inclined to learn about this stuff and just want to use it / solve a problem.

I think that a lot can be said for well written documentation that describes necessary processes to get a desired result in a way that everyone can follow regardless of their prior experience or knowledge.

Security should be the default, but instead a lot of security features are optional things we have dig through docs to set.

TPM support is getting more common, using it should be too. Detected during install? Set it up as part of LUKs during install, and enable a password, and provide option for TANG (both usage or deployment).

fscrypt should be enabled by default and keys set by logical differences of file types. (Yes on top of LUKS). Honestly setup following selinux profiles and per user is a reasonable default. Hardware wrapped keys should be default.

Encrypted memory an option for this CPU? Enable it. Features for multiple key memory encryption? Enable it. Encrypt on a per VM and per container level by default.

Each service should be containerized, connections made explicit (ideally with l7 rules, l4 at least). If a user want to tinker with have a dev mode that opens that service up, with expectation that it’s temporary (track and warn user when active). Each service should run as it’s own non root user.

Each application should containerized. Wayland should be default to minimize shared data. Access by apps should be explicit and user approved and user configurable. Application should never run as root and escalations should be temporary and explicitly approved by the user. Application to the network should be explicit per connection and l7 aware.

MACSec WPA3 pki should be available during install. Wireless WPA3 PKI option should be default on wireless setup. IPSec/Wire guard VPN/Tor should be available option by default on setup. Vlan tagging should be available options on setup.

FIPS or equivalents should be enforced by default. Old encryption methods/cipher/etc should require explicit approval by the user.

Selinux should enabled by default and selinux tagging should be exposed in user applications, so users can choose the security levels, privacy tags (medical or tax docs or etc), or pseudonym access they want.

Sudo should be setup by default for least privileged roles and not god mode access. The combination of those into a single user could look indistinguishable but it should be set and ready for adding users that are limited in scope.

Encrypted backups following the 321 rule (at least 3 backups, 2 different types of media, 1 off site) should be the default and configurable on install. Schedule and triggered backups should be frequently (ideally constantly backup, with snapshot ting being periodic).

Multiple factor logins should be the default. Support for smart card, key fob, OTP, biometric, plus password built-in and encouraged on install.

Number of known CVEs for hardware, packages, and configurations should be tracked and obviously available for privileged users. Hardware missing for full best practices (like TPM 2.0, memory encryption support, etc). Software source should be kept easily accessable to users for remove and modifications. Software should adhere to SLSA build practices, exception explicitly choosen the user.

Systems should be immutable with expectations being explicit to the user and triggering snapshot ting.

DNSSEC and DNSoTLS/DNSoHTTPS should be default and configurable on install.

NTS should be default for NTP configuration. Hardware time sources should be configurable on install.

Applications should be privacy preserving by default (not defaulting to Google for example).

These are just off the top of my head stuff, stuff I had to annoyingly learn and set up myself to harden systems instead of it just being part of sane defauls. CIS bench mark has more controls that should be set.

Snap. The very existence of it.

The fact that there is NO agreed single package standard across distros.

Linux needs a shared API framework for all desktop apps for them to succeed. It’s ridiculous that gnome apps and other apps look different and have different theming conventions. I’d love to get into theming and application building, but I’m so afraid that I’ll waste my time on something that won’t apply to everything. macOS solved application cohesion perfectly.

Bluetooth support can be a mix bag one point my keyboard constantly disconnects for every few minutes likely due to the hardware aggressively try to save power.

Suspending can be 50/50 especially on old hardware. Either you get it back up and running or you will have to forcibly shut it down since it refuses to accept any commands.

1. The lack of a universal application installation method which 98% of developers use. Windows has .exe and it makes it so much easier for developers to release one application which is dead simple for users to install. No instruction manual with different methods per distro. Just double click. This results in less support for Linux in general. Fewer games and applications an drivers with fewer features.

2. Poor backwards compatibility. Yes it results in bloat, but it also makes it much cheaper to develop for and maintain applications, and this results in more developers for Windows. More hardware and driver support. More applications. More games.

It is no mystery to me why developers don’t focus more on Linux support. It’s more expensive. They tell us this. What is so frustrating is that Linux fans are so quick to blame developers instead of focusing inwards and making Linux a more supportive platform for said developers.