Gentoo Linux -- is this the solution to the Winux problem?

I have written elsewhere of the tendency of modern Linux to become increasingly like Microsoft Windows in key areas. A mainstream Linux distribution, like Ubuntu or Fedora, is as easy to install and use as anything from Microsoft. Easier, perhaps, because of the huge range of hardware that Linux supports out-of-the-box.

This simplicity comes at a cost: a modern Linux installation with a graphical desktop is huge and complex, and can be slow on anything but the latest hardware. Moreover, the uptake of frameworks like systemd and pulseaudio has made Linux much more monolithic than it used to be: a few tightly-coupled components, rather than a myriad of independent ones. Systems like the Gnome desktop are not easy to understand, and their internal operation is not well-documented. And yet, Gnome is essentially the default in all mainstream Linux distributions. You have to work really hard to use anything else. And it's hard to use it without systemd.

So, for better or worse, we are witnessing 'Winux' -- the forcing of Windows-style development and marketing practices onto Linux. It's probably no coincidence that Lennart Poettering -- the original creator of systemd and pulseaudio -- now works for Microsoft.

Winux is a phenomenon because of commercial players who have an interest in making Linux easier for a casual user to install and run. In principle, that's a laudable goal. But I see no evidence that Windows users are going to migrate to Linux, however easy it is, and however bad Windows gets. If the current state of Windows isn't enough to drive users over to the Linux camp, I can't imagine how bad it will have to get. Winux is a problem because it discourages enthusiastic Linux users, whilst being unable to encourage casual ones.

One of the appeals of Linux for many people has always been that it doesn't have stringent hardware demands, unlike (say) Windows 11. While it's true that the Linux kernel is comparatively undemanding, the full-fat Gnome desktop is anything but. Maintream Linux is moving in the same general direction as Windows -- increased complexity, reduced flexibility, and reduced support for hardware that wasn't made this year.

For my Raspberry Pi systems, I build the Linux installation completely from scratch, using a bunch of incomprehensible scripts and custom code. I've considered doing the same for my PC-type computers, but it's a heck of a job.

So I was relieved to discover Gentoo. It's been on the periphery of my awareness for years, but I've never gotten around to using it seriously until recently. Gentoo is not really a Linux distribution in the conventional sense but, rather, a framework for building a Linux installation from source. That's generally not as horrific as it sounds, and the Gentoo package management system takes care of most of the routine aspects of the build process. Still, Gentoo is often described as 'hard', even by longstanding Linux enthusiasts.

Building from source opens the possibility of creating a much more custom Linux. If you don't want to use pulseaudio, setting (in theory) a single system configuration variable will exclude support for it from executables as they are built. Things aren't that simple in practice, as I'll explain in due course. Gentoo packaging seems to be more fine-grained than that of other distributions: it's possible, with effort, to control the 'dependency bloat' that is increasingly a side-effect of software packaging methods.

In addition, Gentoo's minimal base install -- before adding any packages -- is much smaller than the so-called minimal install of most other distributions. Quite a few software components that are almost universally used are not included. FUSE (filesystem in userspace) libraries are not part of the base install, for example. This isn't a problem when installing Gentoo packages, because the dependency management system will take care of installing the necessary libraries. But it's harder to install 3rd-party binary software on Gentoo than on other distributions, because it's base install is so, well, basic. But you can't have it both ways, and it's easier to add software to a small installation, than to remove it from a bloated one.

I'm sure there are benefits to the Gentoo maintainers in distributing packages as source. It isn't necessary, for example, to find storage for hundreds of gigabytes of compiled binaries, for dozens of different platforms. Since a few months ago, however, Gentoo has gone a bit more mainstream, and started to include pre-built binary packages as well as source packages. While this does speed up the installation process a bit, it's not as effective as you might hope. Again, I'll explain why later.

Gentoo packaging

For source-based packaging, what the Gentoo repository provides for a particular package is some metadata and build scripts. The scripts will obtain the source from it's usual location -- GitHub, for example -- then build it using options set on the user's system. The metadata will include dependencies, as is the case with all Linux distributions.

However, Gentoo's dependency management system is comparatively complex. In a binary distribution like Fedora, there are binary dependencies between packages. That is, installing one package may cause the package manager to install others that it needs. But Gentoo has build-time dependencies as well. So, for example, the vim editor has a dependency on git. That's not because vim actually uses git at run-time -- rather, installing vim requires installing git to download and build vim. In fact, there aren't just runtime and build-time dependencies: there are at least four different classes of dependency for every package. It's difficult to understand how the packaging system works without installing a Gentoo system; but getting it wrong is likely to make the installation difficult.

Installing Gentoo, in a nutshell

Gentoo does not have a point-and-click installer, which makes it unique among contemporary Linux distributions. Instead, to install you'll need to boot Gentoo from a DVD or USB stick, which gets you a command prompt. From there, you can set up the system using a bunch of commands. Many, many commands, as it turns out.

This process is pretty well-documented. Better still, the Gentoo documentation goes into all the technical details. It explains the various PC-type boot processes in detail, how you might partition a hard disk, and how to set up the compiler for optimization. The documentation explains how an EFI boot works better than anything else I've seen -- it's worth installing Gentoo just to learn about this.

But there's no getting away from the fact that Gentoo installation demands a high level of Linux expertise, despite the thoroughness of the documentation. It's written for experienced Linux enthusiasts, not for casual users.

The 'Gentoo handbook' goes step-by-step through the installation only as far as booting to a console. Beyond that, you'll need to seek out specific guidance for common tasks, like installing an X server. Then you'll have to find further information if you want a full desktop, or printer support or, well, pretty much everything.

I do want to stress that the documentation is pretty good. But there's a lot of it, and finding your way around it can be a challenge.

Portage, and USE flags

Portage is Gentoo's package management and build system. In a sense, Portage is Gentoo. So understanding it -- which is a daunting task -- is key to making good use of Gentoo. Portage is exposed as a bunch of command-line utilities, most notably emerge. So, in Gentoo jargon, installing a package is described as 'emerging' it.

The Portage build and installation process is governed by 'USE flags'. These can be set at the system level, or for individual packages. A USE flag controls some aspect of the system's capabilities. The flag -systemd, for example, means that packages should be built without support for systemd (and, crucially, without a dependency on systemd). The USE flag X means, in general, that X support should be included in packages.

USE flags work at build/installation time. If you change them at the system level after packages have been built and installed, that's potentially a problem, because these flags will be out of sync with the installation. There are ways to rectify this, but they can be somewhat time-consuming.

USE flags don't just control how packages are built: they control the dependencies that exist between packages. The centrality of these flags can hardly be overstated.

Another interesting feature of Gentoo that I've not seen elsewhere is fine-grained control over licensing constraints. Some people prefer to be sure that they're only installing open-source software, with some specific licence terms. Some will be willing to install proprietary binaries, provided that the license terms are clear. Some people will be happy to install anything, so long as it works. Every Gentoo package has its licensing terms in its metadata. Packages will be blocked ('masked' in Gentoo jargon) if the licence conflicts with the user's constraints. Distributions like Fedora and Ununtu do expose the licence terms -- legally, they have to -- but their package installers don't take these things into account in the way Gentoo's does. Other distributions implement a measure of licence control by having different repositories for software with different legal categories. But, so far as I know, no distribution but gentoo has licence control as a first-class feature of its package management process.

As impressive as this seems, I found that I got fed up with specifying licence constraints on a package-by-package basis, even though I could. So I ended up making a system-wide setting to the effect that emerge could install anything that was flagged as redistributable, regardless of the specific licence. As keen as I am, in principle, to run only open-source stuff, some hardware requires proprietary, binary drivers. Network adapters are a particular case in point.

Why binary package distribution isn't a silver bullet

If you want a highly customized Linux, Gentoo provides -- in principle -- a way to get it. The existence of binary packages suggests that we might be able to get the benefits of this customization, along with rapid installation. The problem, if that's the right word, is that the Portage system is not really compatible with binary distribution. Not at its core.

Each binary package in the Gentoo repository will have been built to be compatible with specific USE flags. If you try to install a package where these flags are not compatible with your settings, Portage will generally build from source. So, for example, I have USE=-pulseaudio because I won't be running pulseaudio on my system. If I try to install the Firefox web browser, which is as simple as running emerge firefox, I won't get the binary version. Why? Because it wasn't built for a system without pulseaudio.

I can still allow emerge to build Firefox from source, but that will take a minimum of two hours. On some systems it could take days. And if the build fails a couple of times first, for whatever reason, it's going to take a long, long time just to get a web browser.

I should point out that, usually, the build does go smoothly. When it doesn't, it's a real nuisance.

Behemoths like LibreOffice and Thunderbird are as problematic as Firefox. Building an X desktop environment -- even a relatively lightweight one like Xfce4 -- is not a five-minute job, either.

Gentoo's build system allows me to take a chance, and force emerge to install the binary, even if it seems to be incompatible. Sometimes this works, and sometimes it doesn't. It seems to be worth trying, at least, because you can always revert to the source build if the binary isn't satisfactory.

Don't get me wrong: the emphasis on building from source means that it's possible to get a very focused, highly-optimized Linux, with few unnecessary dependencies. But it takes a certain amount of experience, and a few failed installs, to get to that point. The existence of binary packages doesn't really change that and, in practice, it doesn't always offer the improvement in installation time that we might hope for.

Still, once you've arrived at a successful install, and built everything that needs to be built, you can take the compiled binaries and build a repository specific for your needs. The initial set-up of such a repository takes a bit of work, but it's likely to be repaid in the long term.

Gentoo and systemd

Gentoo is one of the few contemporary Linux distributions that don't force the use of systemd down our throats. It has a well-documented alternative, in the form of OpenRC. Of course, OpenRC is just a service manager -- as it should be. You'll still need other subsystems to take care of all the tasks that systemd has absorbed into itself. This is unlikely to be a problem for an old-school Linux user, and the process is well-documented.

What can be a problem is the way that systemd has infiltrated itself into so many other Linux packages. It's very difficult to use the Gnome desktop without systemd -- Gentoo has created stubs for some of the services that Gnome relies on, but Gnome-with-OpenRC support has a patchy history in Gentoo. Other popular components, like pulseaudio, favour the use of systemd, and they have long tentacles. Almost every piece of Linux software that can make a sound has some level of dependence on pulseaudio.

What this all means is that, if you're not careful, you'll end up with systemd, or parts of it, even if you don't want it. The Gentoo documentation describes in detail how you can prevent systemd being dragged into the installation by means of its sprawling dependencies. But it concedes, all the same, that sometimes it just won't be possible to install what you want, without systemd. You'll have to look for an alternative.

That's not the fault of Gentoo -- it's just a consequence of the way Linux has been manipulated by commercial interests that have a stake in making Linux more like Windows,

As things stand, even Gentoo users can't completely avoid the 'Windowsification' of Linux. The udev daemon, for example, which is more-or-less essential for a graphical desktop, has been completely absorbed by systemd. So you'll likely have to run systemd-udev even if you don't use any other systemd component.

Gentoo and Wayland

Wayland is the graphical interface system that is slated to replace the ageing X/11. At present, Wayland is compatible (or should be) with X applications, but there are advantages to applications in using Wayland directly. Of course, these applications won't work on a system without Wayland.

I don't know how I feel about Wayland -- it has some huge advantages over X, but uptake has been, well, variable. I have no objection to it on principle -- it isn't bad to its bones the way that systemd is. But it remains a problem that a lot of older hardware doesn't support it, and most Linux distributions are moving towards supporting nothing else.

At present, I haven't found this to be a problem in Gentoo -- I've not found anything I need to use being unavailable for lack of Wayland support.

Gentoo and updates

On the whole, Gentoo does not have specific release milestones. A Gentoo installation doesn't really have a version number in the way that, say, Fedora has. Instead, each package is updated as needed, and new versions are available immediately.

There are benefits, but also costs, to this approach. The main benefit is that we always get the latest available bug fixes and security patches, without waiting for a new release. The cost, however, is that there's no specific Gentoo platform version that 3rd-party, commercial software can certify against. It's unlikely that we'll see Gentoo hosting a production on-line banking service but, to be fair, very few Linux distributions are used in such a way.

Gentoo support

So far as I know, nobody is offering commercial support for Gentoo. Relatively few organizations offer paid Linux support and, so far as I know, all those that do focus on mainstream, heavyweight installations with a high degree of Winux.

Gentoo has a web forum for discussion and ad-hoc support. In practice, I've found that community support helped in about 10% of my problems with Gentoo; the rest I had to work out myself, or live with.

The very flexibility of Gentoo means that it's hard to support. Everybody running, say, RHEL 9 is running essentially the same software, configured in essentially the same way, on similar hardware. It's a good environment on which to provide commercial support, but even the public forums are quite effective, because everybody tends to see the same problems.

Gentoo users need to be pretty skilled to start with, and then be willing to solve most of their problems themselves.

Gentoo derivatives

Gentoo has been around a long time, and has given rise to a number of spin-offs. Broadly, these attempt to make Gentoo easier to install, while still retaining some of the benefits that make Gentoo stand out. Of particular note is Funtoo, which is a project started and run by Danny Robbins, the originator of Gentoo. Funtoo has a larger amount of binary distribution, and is somewhat more opinionated than Gentoo. So, while systemd and Wayland are options in Gentoo, they are unsupported in Funtoo. Users are expected to run the (binary) production kernel, rather than building their own. Low-level fiddling using USE flags and similar techniques is discouraged.

The problem with these Gentoo spin-offs is that they partition what is already a small user/developer space into even smaller ones. It's worth bearing in mind that, with an estimated 100,000 users, Gentoo itself accounts for only about 0.2% of Linux users. You can't divide this user population up very many times, before you end up with a non-viable number.

Moreover, all the Gentoo derivatives appear to be maintained by a few, very active users. While these folks deserve credit, I'm not sure that this is a sustainable management model.

So where does this leave us?

Gentoo Linux remains a good choice for Linux enthusiasts who would otherwise be building their own Linux installations from scratch. As difficult and time-consuming as it can be to install Gentoo, it's still quicker and easier than hacking up a complete Linux desktop system using scripts.

Even if you don't need this level of custom control, there's a place for Gentoo among Linux users who just want their software highly optimized for their hardware.

If you don't want systemd and everything that goes with it, I think the recent inclusion of binary distribution won't make Gentoo any quicker to install. Few of the binary packages are going to suit your system. I'm finding that about 60% of the software I install has to be built from source, and it can take a long, long time. Gentoo isn't the Linux you want, if you want it today. It's also not the Linux you want if you just want to try Linux. I'm fairly sure it's only going to appeal to long-time Linux geeks like me.

To be frank, on most computers I use, I don't need Gentoo. I like its reduced size and improved efficiency, but I don't need them. That's because most contemporary computers are hugely over-resourced for desktop work. As inefficient as the Gnome desktop is, it's not going to be a problem on the computer I use to simulate a production Kubernetes cluster with sixteen concurrent virtual machines. It offends me at a visceral level that a modern desktop computer needs at least 16Gb of RAM and hardware-accelerated graphics rendering just to run a word processor. But, frankly, it doesn't offend me enough to spend four days installing Gentoo on hardware like this. Fedora and Ubuntu work fine, even though I have to grit my teeth a little over systemd.

Ironically, where Gentoo offers the greatest benefits is where it's hardest to install. I've found that Gentoo really flies on my 2012-era Lenovo 520, but it took four whole days to install.

A four-day install isn't so bad, if you only have to do it very occasionally. But it's always possible to mess up an installation, so badly that reinstallation is the only recourse. At least, it's always possible for me to do so. Moreover, the Gentoo documentation is very clear that, if a system isn't updated for more than a few months -- and updates can be very time-consuming -- it will be difficult, if not possible, to update without a full installation. This is true for Fedora and the like as well; but a complete re-install of Fedora takes me less than a hour these days.

Closing remarks

I like Gentoo and, more importantly, I like what it stands for: a way to install a version of Linux that is efficient and productive, even on under-resourced hardware. I like the fact that it hasn't been entirely swept up into the world of Winux, although a measure of Windowsification is apparent even in Gentoo.

There are, at present, enough users and enough developers to make Gentoo a viable proposition. I hope that remains the case, because I don't like the direction most other Linux distributions are headed.