I’ve linked to quite a few posts by OpenBSD developer Solène Rapenne on OSNews, mostly about her work for and knowledge of OpenBSD. However, she recently posted about her decision to leave the OpenBSD team, and it mostly comes down to the fact she hasn’t been using OpenBSD for a while now due to a myriad of problems she’s encountering. Posts like these are generally not that fun to link to, and I’ve been debating about this for a few days now, but I think highlighting such problems, especially when detailed by a now-former OpenBSD developer, is an important thing to do.
Hardware compatibility is an issue because OpenBSD has no Bluetooth support, its gamepad support is fractured and limited, and most of all, battery life and heat are a major issue, as Solène notes that “OpenBSD draws more power than alternatives, by a good margin”. For her devops work, she also needs to run a lot of software in virtual machines, and this seems to be a big problem on OpenBSD, as performance in this area seems limited. Lastly, OpenBSD seems to be having stability issues and crashes a lot for her, and while this in an of itself is a big problem already, it’s compounded by the fact that OpenBSD’s file system is quite outdated, and most crashes will lead to corrupted or lost files, since the file system doesn’t have any features to mitigate this.
I went through a similar, but obviously much shorter and far less well-informed experience with OpenBSD myself. It’s such a neat, understandable, and well-thought out operating system, but its limitations are obvious, and they will start to bother you sooner or later if you’re trying to use it as a general purpose operating system. While it’s entirely understandable because OpenBSD’s main goal is not the desktop, it still sucks because everything else about the operating system is so damn nice and welcoming.
Solène found her alternative in Linux and Qubes OS:
I moved from OpenBSD to Qubes OS for almost everything (except playing video games) on which I run Fedora virtual machines (approximately 20 VM simultaneously in average). This provides me better security than OpenBSD could provide me as I am able to separate every context into different spaces, this is absolutely hardcore for most users, but I just can’t go back to a traditional system after this.
↫ Solène Rapenne
She lists quite a few Linux features she particularly likes and why, such as cgroups, systemd, modern file systems like Btrfs and ZFS, SELinux, and more. It’s quite rare to see someone of her calibre so openly list the shortcomings of the system she clearly otherwise loves and put a lot of effort in, and move to what is generally looked at with some disdain within the community she came from. It also highlights that issues with running OpenBSD as a general purpose operating system are not confined to less experienced users such as myself, but extend towards extremely experienced and knowledgeable people like actual OpenBSD developers.
I’m definitely not advocating for OpenBSD to change course or make a hard pivot to becoming a desktop operating system, but I do think that even within the confines of a server operating system there’s room for at least things like a much improved and faster file system that provides the modern features server users expect, too.
Who knows what the future will bring but i agree that currently GNU/Linux is it, the best option out there, for everything.
It’s very important to be flexible and pragmatic, and give credit where credit is due. Things change, and the OS you love may change, too. Moreover, there’s hardly ever a one-size-fits-all approach for all situations once you start your operating system journey.
I migrated to Windows 2000 from 9x and accepted I could not run many of my games (especially at the beginning) and games mattered to me, because I was a teenager. I just never managed to get used to all the instability and Windows 2000 required a full reinstallation only 2x per year instead of once every two months for me.
Then I had access to a G4 Mac at work (early 2003) and, even though I enjoyed the OS 9 GUI, it just crashed too often for my taste. Once I was allowed to upgrade to Mac OS X, I was sold. The power of UNIX with the broad application base available to the Mac, with Mac OS X being my daily-driver from 10.3 to Mojave.
For some time, I also ran Debian, but I started getting very irritated by how often basic things changed in Linux and the first time I wrote ifconfig and got a “not found” error… ugh, I just can’t with relearning the basics every few years, and I migrated some work to FreeBSD.
Meanwhile, I became more privacy-conscious, so I started using Qubes OS on a corebooted ThinkPad X230, and this is the machine I use for almost everything that requires an online account, and a FreeBSD ThinkPad W530 for the rest (with the W520 keyboard, of course). And Solene is right, Qubes OS is very slow.
However, I kept the Mojave 2009 Mac Pro for some serious work – I lost a few precious photos due to bit rot and became a fan of the ECC+ZFS combo (my servers run ZFS, and I operate with a remote home folder).
With the complete dumbification and iOSzation of Mac OS, now I want to get rid of the Mac Pro, so I got a ThinkPad P1 with ECC memory and, well, the battery life is half of what it is under Windows (the Intel GPU drivers somehow don’t let the CPU sleep properly). Credit where credit is due, Windows is a decent virtualization platform. My ThinkPad has 2 NVMEs with ZFS, 64GB of RAM and I can either run FreeBSD on the bare metal or boot it under VirtualBox in Windows. I save my data in the mirrored ZFS partition, run most of what I want to use virtually, and it is ok. I hardly ever need to see the Windows desktop and I get 90% of what the Windows-only battery life is. It was painful to make Windows leaner and remove all the telemetry rubbish I could, but now it works ok – and I get to game if I want to. I could have gone the Linux way and have better support than FreeBSD, but a few things in Linux irritate me.
My servers saw Debian, CentOS and now they are on FreeBSD.
If I really could, and if the system had a bit more security built-in, I’d love to daily-drive Haiku. It’s the best UI out there.
– So, if you need full compartimentalization of work, high privacy or under direct personal surveillance threat, go Qubes.
– If you really don’t care about privacy and want a simple out-of-box experience, Windows or Mac are just fine.
– Linux, as something like Ubuntu or Fedora, are great if your computer are not supported by the latest releases of Windows or Mac, or if you want to keep an older computer running. It is also the best alternative for modern hardware, as the BSDs tend to take longer to catch up in terms of hardware support.
– The BSDs are great options for older hardware, including laptops, and you can get an incredibly lean fully functional system. My W530 goes for 5+ hours on FreeBSD, way longer than under Windows. The system starts with less than 10 kernel modules loaded!
The sad thing, to me, about Windows, is that the bloat that Microsoft forces on it ruins a very fine operating system. Dave Cutler designed a beast!
And about the Mac, it is just sad that Apple is going the way of the privacy nightmare and the system is also very bloated of services I have no idea what they are doing, and getting more and more locked down.
For the OSNews geek, well, I guess virtualization will give the best of all worlds at all times. =)
And we need to remember that things change – Linux can get worse, Windows can get better. Keep your data portable and adapt!
> The sad thing, to me, about Windows, is that the bloat that Microsoft forces on it ruins a very fine operating system. Dave Cutler designed a beast!
I really wonder if MS devs are running some kind of an engineering edition of Windows at home, or if they really get to cope with all the same crap as we do.
It’s already so frustrating as a user, I can’t imagine what it is to see your fine work being turned to trash every 2 weeks when some dumbass product manager comes up with a new piece of crap he wants to force down the throat of 1.4 billion people, just because he feels like doing so.
I imagine if such a Windows version exists, the users would view it as the natural continuation of Windows 2000 with flatter graphics, and any contact with consumer grade Windows is incidental and incites “what the fuck” and “wow, people deal with this shit?” and also the only real complaint is the lack of sloppy focus.
They absolutely are, it’s called Windows Enterprise. Anyone who works for Microsoft and any independent developer who joins Microsoft’s developer program, has access to Enterprise. It is literally a bloat-free, extremely basic installation of Windows. No Microsoft Store, no Xbox bullshit, nothing that could be considered a consumer “feature” is installed.
Sadly, us plebs don’t have (legal) access to the Enterprise version unless our employer has a contract with Microsoft. The company I work for is too small to have a contract so we bought Windows Server standalone that comes with 25 user CALs, and we use the Pro version of Windows that comes with the workstations we buy. Whenever I set up a new workstation I have to go through and clean out all the junk and turn off/down all the telemetry and other annoying settings. This is streamlined with scripts but it’s still annoying. I wouldn’t have to do any of that with Enterprise, but we are a small business.
No they’re not.
I mean, I’m talking about the actual Windows devs.
I really doubt the guys running a freshly compiled OS at work are going to run the 5 years old enterprise version at home.
But also, there’s no way they’re running the consumer shitty version either. Unless of course there’s a magical kill switch for all these “features” only them know about.
It is exactly as I have said before. The main thing you get with these alternative Unix based OSes as a desktop is Linux, but worse.
The problem is that these OSes are spread too thin, with only a few main devs per project. FreeBSD was quite a lively project in the late ’90s, for example, but it’s almost crickets now. It’s even worse for NetBSD and OpenBSD. All of them together, they probably have about 50 serious devs working on these projects (I’m probably generous). Consider that Be, Inc. had 90 engineers at its peak, working for as single OS.
To me it’s obvious what needs to be done, and it has been obvious since the mid 2000s, when DragonFly BSD happened: all these forked projects need to come together under a new BSD umbrella, setup a tri-andry for executive decisions (e.g. the main leaders of the 3 BSDs), and bring all these devs together. Bring with you the best parts of each project (e.g. security from openbsd, portability from netbsd, extensive support from freebsd etc).
Equally as important is that the new project needs to target desktops and mobile, not just servers. Otherwise, they will never get new blood in. For someone to fall in love with a project, they’ll need to use it, and few people are running servers (and the ones who do, they prefer Linux).
Or, they can stay forked, until it’s like the Gnu Hurd situation where it’s just 1 person working on it, and it has the most mysterious bugs you’ll ever see (e.g. X11 working fine, rebooting, and not able to run X again — weird stuff).
Pretty certain the OpenBSD devs have a very different focus than the others and they love to rip things out they don’t need or considered not secure.
They don’t work well with others.
We can look back to the reason they forked in the first place. The NetBSD bureaucracy tends to crush egos.
Regarding the OP, billions of corporate dollars being poured into the biggest open source projects has warped people’s perspectives of what it’s possible to achieve as a non-profit project. Sacrifices will be made.
I think the problem with cooperation is that most open source projects happen because someone wants something that they would use themselves – so when people disagree, you start getting forks. You don’t want to do unpaid work of love for something you don’t love. Why would you compromise when you are doing it “for fun”?
The fact that we get functional operating systems out of the work of “50 serious devs” shows, to me, how much bloat there is in the world of software today. Of course, there are others writing drivers and applications, but you get the point.
Linux got where it did because there’s Linus with supreme veto power. The day he is gone, I believe we will end up with HP-LINUX, DELL-LINUX, MSFT-LINUX, just like we had during the UNIX-wars.
“Linux got where it did because there’s Linus with supreme veto power. The day he is gone, I believe we will end up with HP-LINUX, DELL-LINUX, MSFT-LINUX, just like we had during the UNIX-wars.”
I think the split between kernel and Linux distributions prevents this from happening, for example let’s take Microsoft already has kernel developers and Linux kernel is pretty accepting of code if you are professional and can work together with others. And they have their own Linux distribution, AzureLinux..
The unix wars followed completely different set of dynamics FWIW.
Just give me HammerFS and Signal or Session messenger and I’ll switch back to OpenBSD.
What you propose is simply not realistic. The BSD’s are all unique operating systems with decades of work separating them.
Some quick stats, pulled from recent commit history/openhub.
monthly people committing, ~October 2024
• OpenBSD: ~84 (ports + kernel + base)
• FreeBSD:~113 (kern + base)
•~226 (ports) w/ overlap
That’s a lot more than “50” people, not including NetBSD and other extant *BSD projects. Despite what people think, these projects are still gaining contributors year after year.
The problem is that even the biggest (FreeBSD) has orders of magnitude less active developers than linux.
If anything 113 is tiny, in terms of active core developers for FreeBSD, given how some major commercial products (PS5 et al) are based on it.
You got a point but it will never happen. The best case would be if we could somehow could get the best bits of OpenBSD, NetBSD and DragonflyBSD – and maybe Illumos too – and implement them in FreeBSD or a newly created soft-fork of FreeBSD (with a Debian Ubuntu like relationship).
I just checked. OpenBSD removed SoftUpdates on their File Systems in 7.4. No definitive replacement has been announced. O_o
Running check disk on a full 8TB HDD (or rebuild raid5 set) must be an experience to live once.
Filesystem was already a lock-in a few years ago. I moved to Debian after a decade and migrating 6TB of data took weeks because the system would always crash at the 2TB mark if writing directly, so I had to use NFS which was painfully slow. Just that portion delayed the transition for at least one release cycle.
This definitely a case where the OS doesn’t fit the user anymore. I’ve been using OpenBSD as my primary system for something like ten years. It does run hot and draw power, its something I’ve noticed. The filesystem does need some TLC.
Haven’t read it yet, but I’m glad you posted it. I think posting things like this is valuable as well, especially when they’re someone’s reasoned, thought out, thoughts and opinions on why they took a specific action around OSes and software and the like. More good opinions the better.
To a large extend, Solène Rapenne is switching from OpenBSD to a manufacturer-supported OS. Unfortunately, most of the drivers at OpenBSD are inverse-engineering made, this leads instabilities on newer hardware and this cannot be avoided. Fortunately, Linux got traction enough some years ago to catch the attention of the hardware industry.
Anyhow, just discussing how inconvenient is OpenBSD on the desktop is already a big success. 10 years ago it would be only a bad joke.
Unfortunately the performance delta between the BSDs and Linux on the same hardware keeps getting worse, so they will continue bleeding more and more users/developers. So the chicken and egg cycle will keep getting worse and worse.
it’s what it is.
I start to be worry reading the comments here.
There was a time when gnu/linux performed bad in several cases. Not talking about instabilities on new hardware. But there were a community supporting not the technical supremacy but the freedom of choice , code interoperability and open standards. From the comments here (and the article quoted) it looks like only technical parameters are worth. Where would be Linux if that would be the case at the beginnings of free software?
But Linux is here. Free software is here. That’s why BSD don’t bother to many people. It’s not SO important unless you want something that corporations can lock and don’t give back.
13 outright lies in that article.
Can you be more precise if you suggest that OpenBSD dev lie?
This article gives an impression that doesn’t tie in with my own experience, which is that OpenBSD is a perfectly viable daily driver as a laptop/desktop OS, though that isn’t its main purpose.
Some more advanced users may find that it doesn’t suit them, either due to missing functionality, lack of hardware support, or needs to run software that isn’t or can’t be ported. I certainly respect Solene’s decision to move on to other platforms, certainly she has fairly advanced requirements and losing data in search of them must be very frustrating, but I hope that she still comes back to using OpenBSD in the future. It has many benefits, including relative simplicity, security focus, ease of installation, and believe it or not, good hardware support on laptop platforms out of the box.
Perhaps I should explain that last claim a little.
I am a long-time Linux user (25 years plus), and used FreeBSD as a server in production for several years in the early 2000s. I have been finding Linux a little frustrating as its complexity grows in recent years, so I decided to try OpenBSD on a variety of machines, in the hope of a simpler life.
So far, my experience has been as follows:
– Two Thinkpads worked out-of-the box with a standard install, with full hardware support and suspend/resume. (I wasn’t successful in getting suspend/resume working on FreeBSD on either, though Linux also worked out-of-box).
– An older Panasonic Toughbook (2012) failed to Resume after suspend, until I figured out the issue was with the TPM, and the BIOS wasn’t properly disabling it when I tried to work around it. A bit of boot-time configuration, and now, everything works.
– A Toshiba NB200 Atom-based netbook- everything works (i386)
– Toshiba R930 Ultrabook – everything hardware worked ‘out of the box’.
– Japanese-market Panasonic Let’s Note. This was a big ask, because these slightly odd machines are not common outside Japan. The SDCard slot didn’t work out-of-the-box. Fixed with an existing patch that has since been committed. Touchpad scrolling didn’t work, because the item was being identified as a mouse instead of a Synaptics. Putting debug statements in the driver, I traced it to a lack of legacy protocol support in the firmware, and made a tiny patch to the Keyboard/Mouse driver. Now it works.
Regarding performance, I haven’t noticed any difference in practical use to my Linux machines, though a benchmark might show some. Battery life is acceptable, Maybe power management isn’t as advanced as Linux, but the real-world differences aren’t huge and not a show-stopper.
Regarding the filesystem, I haven’t found UFS to be unreliable, or slow. Debugging Resume on the Toughbook, I must have had to force-shutdown fifty times. Fsck fixed any resulting filesystem inconsistency after reboot every time.
The configuration steps from ‘standard install’ to ‘my standard system setup’ are not many, easy to do and to learn, and are certainly no more complex than under Linux. Command syntax for tasks like network setup are simpler and more logical.
Things that might discourage some other potential users:
– VM support is functional and easy to set up. But VMs are single-core and without graphical display, at least for now. Workarounds are possible, using Xephyr or VNC, however. No Windows VM support, for now. OpenBSD or Linux only.
– Multibooting is possible, but not encouraged, due to all the varied and complex issues that may result. Unless you really know what you’re doing with booting, it’s best to give OpenBSD a whole disk. I was able to make a quad-boot machine without too much trouble, just to prove it could be done.
– Some Linux-only software hasn’t been ported yet, particularly if heavily dependent on things like Systemd, though Gnome and KDE have been done. I haven’t tested them, as I prefer to use lightweight Window Managers instead. GTK/Qt applications run just fine.
– Some proprietary hardware, even that supplied with barely-functional Linux drivers, may not work. This isn’t always a show stopper, however. I was able to set up a non-Wacom USB tablet with a bit of careful Xorg config, and no “driver”, and it all works fine including pressure-sensitive drawing.
– there’s no NetworkManager, GUI applet, or similar. You have to configure networks by editing a configuration file. Easy to learn though, and once done, the functionality is the same.
– Technical help is via Mailing Lists, not forums. You are expected to provide technical details of issues beyond ‘X doesn’t work for me, Help!’ If you are used to that, it’s fine. It helps to already know how to wrangle such things as Xorg, fonts, Themes etc.
– The system documentation is by Manpages (high quality and accurate in my experience) and the fairly terse website FAQ (same). This might not accord with what you may have experienced elsewhere.
If you are wanting to give OpenBSD a try, by all means do so. I am glad I did. You are likely to be pleasantly surprised.
If on a laptop, it’s best to start with a Thinkpad or common HP model, though it’s not impossible to overcome incompatibilites on other machines, such as Wireless chipsets, by inexpensively swapping the card for a supported model.
I’m looking forward to learning more about OpenBSD in the future, and feeding back the knowledge gained to the wider community.