Hacker News Reader

Comment by insulanian | Hacker News Reader

Honestly I don't think FreeBSD should focus on desktop story at all. Resources, as limited as they are, should be spent on making FreeBSD an obvious and compelling choice for the servers.

Although FreeBSD looks cleaner and more consistent than Linux, it's not really clear why should one use it over Linux.

I hope the team will define strategic goals and try to develop some competitive advantage, even if it means focusing only on certain niches in the beginning. E.g. encouraging development of the OS modules in a safe language like Rust would probably lead to more stable system less vulnerable to exploits, and make it a more attractive choice for mission-critical deployments.

View on HN

At least for me, being able to use Linux as daily driver on my desktop was instrumental at making me more proficient. At that point using it in the server became natural. Nothing beats the exercise of being able to use the same environment for everyday computing and for running "serious" services.

I guess a good desktop experience has the same utility to the freebsd ecosystem.

I've had the same experience. Used FreeBSD as a server OS long enough to switch to it as my main desktop OS. In return I've learned a lot more that has made server administration better.

I'm not sure I'd apply this to Windows, personally.

Companies started to use Windows NT servers in the 90s maybe because they cost less than UNIX servers from HP, IBM, Sun etc but also because they had a lot of technical staff which only knew Windows.

> Honestly I don't think FreeBSD should focus on desktop story at all.

I wish they didn't have to. I wish the open source community would rethink the desktop experience from the ground up so that we can have truly portable desktop environments that speak a common API. Everything is so tightly coupled to the platform and a spaghetti mess of packages and tools that I don't think it can be achieved any way else.

Apple has vertical integration of their OS, so they don't have to worry about supporting the world. But the state of things in open source land right now is a complete disaster. X vs Wayland, compositors? desktop environments? window managers?

For instance, check out the Manjaro doc on switching desktop environments: https://wiki.manjaro.org/index.php/Install_Desktop_Environme...

Why is this such a nightmare? There is an entire section called "The Risks of Using Multiple DEs".

There are 'two steps' for switching to XFCE:

1. Install a basic XFCE environment,

    sudo pacman -S xfce4-gtk3 xfce4-goodies xfce4-terminal network-manager-applet xfce4-notifyd-gtk3 xfce4-whiskermenu-plugin-gtk3 tumbler engrampa

2. Optional: Install and use LightDM, the recommended display manager for XFCE (record comes screeching to a halt -- what is a display manager and why is LightDM recommended?)

    sudo pacman -S lightdm lightdm-gtk-greeter lightdm-gtk-greeter-settings && sudo systemctl enable lightdm.service --force

This is relatively simple when you look at it here, but seeing this x 10 on the linked page where every single experience has a vastly different install and configuration process does not give me warm and fuzzy feelings.

Anyway I got no suggestions on how to improve this -- rant over. I just wish that we'd settle on a really rock solid abstraction and then get back to work rethinking and rebuilding UIs in Linux/BSD. Without vertical integration and/or monetary incentives I do not see this ever happening.

To be frank, setting up different DE, window managers, etc. is one of the few places where linux has little issue. The only problem being themes or having multiple bluetooth managers at the end of it. It's really just as easy as "sudo pacman xfce" and then selecting xfce from the display manager.

The real issue is that you have all those choices, but none of them are as fully featured without caveat as a similar DE for Mac or Windows. And there's the separation between window manager, desktop environment and display manager which is very confusing.

You're inflicting this upon yourself by using Manjaro. Installing a different DE on Ubuntu is either a Xubuntu-desktop metapackage, or a tasksel menu-item, away.

> X vs Wayland, compositors? desktop environments? window managers?

Nothing on this list, besides desktop environments, are things the user ever has to think about, unless they go out of their way to.

> Nothing on this list, besides desktop environments, are things the user ever has to think about, unless they go out of their way to.

It seems you haven’t tried to use Ubuntu as a desktop, especially not outside of command line. I didn’t care about anything on the list but had to make decisions often because, from my point of view, the GUI applications providing basic functionality just didn’t work. The answer was many times “it’s because it’s Wayland” or compositor or whatever. Horrible experience, when I've used (just an example) gpart for years, it worked, then install a new version of Ubuntu and it doesn't. Then I investigate why, because Wayland. And I really never wanted to care about Wayland and compositor. Then I want to record a screen interaction. Should be easy. Again... doesn't work, investigate... "it's because it doesn't on the new compositor whatever"... Etc.

I suspect that those who develop GUI on Linux don't use themselves what they produce, as a lot of GUI is often unusable, or, even worse, was usable once and then completely degrades, even with newer... compositors, or even some other kind of "improvement."

> I suspect that those who develop GUI on Linux don't use themselves what they produce, as a lot of GUI is often unusable, or, even worse, was usable once and then completely degrades, even with newer... compositors, or even some other kind of "improvement."

I'd agree with that. I'd go so far as to say most GUI developers on Linux may never have actually used a GUI in their lives and rely on second-hand descriptions of how they work. Basic drag and drop often doesn't work, clipboards are all over the place, window behavior is super inconsistent... at least PulseAudio, for all its many faults, finally solved the sound server problem.

1) There's no Ubuntu LTS release with Wayland as default display server.

2) Gparted doesn't work, because it wants to run as root. You should use an application, that runs only minimal, necessary parts as root, and the UI runs with your user's capabilities.

3) Wayland compositors do not provide unrestricted access to to composed display for very good reasons. They do provide an access through controlled API. That controlled API didn't stabilize into unified one among different compositors yet, so for screen recording you have to use compositor-specific utility for now. In Gnome-Wayland/Ubuntu, you have one available by default.

So yes, when you clean up the capabilities over time, the old stuff will break. But without breaking it, you cannot develop your system. Apple also breaks compatibility will old apps, even more than most Linux distros.

And yet I can still often run GUI programs written in 1995 on Windows 10 without recompilation. Hell, I can usually run GUI programs written for Windows in 1995 on Linux without recompilation. Don't pretend that compatibility is some kind of impossible dream.

> And yet I can still often run GUI programs written in 1995 on Windows 10 without recompilation.

Did you try it? Often, you cannot run even programs built in 2009 on Windows 10.

> Well, I can usually run GUI programs written for Windows in 1995 on Linux without recompilation.

You can run 1995 Linux binaries in Linux, provided a) you have the libraries they were linked against, b) they don't use direct hardware access ala svgalib c) they are not in a.out format, which was recently removed, due to nobody using it anymore.

> Did you try it? Often, you cannot run even programs built in 2009 on Windows 10.

Yeah, I actually do it all the time. There are some programs that don't work, particularly games, but for the most part it works fine.

> You can run 1995 Linux binaries in Linux, provided a) you have the libraries they were linked against

Including glibc, and take care to jump through the appropriate hoops to ensure the linker can find them because chances are they aren't in your repo.

> It seems you haven’t tried to use Ubuntu as a desktop, especially not outside of command line.

I'm typing this on Ubuntu.

> The answer was many times “it’s because it’s Wayland”

I don't know how that is an answer you've come across as a solution on Ubuntu, because Ubuntu does not use Wayland unless you go out of your way to use Wayland.

> Horrible experience, when I've used (just an example) gpart for years, it worked, then install a new version of Ubuntu and it doesn't. Then I investigate why, because Wayland

I've used gparted for more than a decade and never ran into an issue with it. Not sure why you have a problem with Wayland when Ubuntu doesn't ship with Wayland support.

> And I really never wanted to care about Wayland and compositor.

If you were using Ubuntu, you never had to care about Wayland, unless you inflicted that problem upon yourself by going out of your way to use Wayland.

I respectfully disagree.

> I wish the open source community would rethink the desktop experience from the ground up so that we can have truly portable desktop environments that speak a common API.

Every distribution does it their own way. The correct comparison of FreeBSD is not Linux; it is a Linux distribution such as for example Ubuntu, Manjaro, Arch, Debian, ...

These distributions each have a relevant market share, making the comparison fair.

Sure, there's overlap. If you want an OS which does it differently, consider having a look at NixOS.

> For instance, check out the Manjaro doc on switching desktop environments [...]

> There are 'two steps' for switching to XFCE:

> sudo pacman

> [...]

You're not supposed to execute commands which you don't understand. You're not supposed to execute sudo commands which you don't understand. Manjaro is supposed to be a user-friendly Arch version, so that is at odds with using CLI if you ask me.

I agree with you about the monetary incentive. Thus far, it does not appear anyone has succeeded, apart from Apple with iOS which is based on macOS which is based on FreeBSD. Jolla's Sailfish is struggling. Librem 5 with Purism's PureOS is delayed. Perhaps PinePhone?

Never going to happen, GNOME and KDE are nice, but their frameworks experience still fails short of other desktop environments, even after all these years.

They depend on too many moving pieces underneath them, and then there is the community hate how bloated they are, and how wonderful it is to reduce the experience to the twm days.

I don't necessarily agree with this. Running on servers was not a priority for Linux but it took off in servers. I think part of this is because a lot of people tried Linux at home, liked it a lot, and then brought it to work. Kind of like how iPhones got adopted in big companies.

I am worried that if FreeBSD completely ditches the desktop side completely, it will become one of the OSes that the next generation of developers would not even encounter unless it happens to be used at their jobs.

Naturally it took off on servers, GNU/Linux is primarily used as free beer UNIX, and it had corporate backing, including UNIX vendors that saw it as a means to cut down costs.

The BSDs were in legal problems back then, plus with their license, many of its users never contribute anything back.

Except for SGI and NeXT, there was hardly any desktop culture on UNIX land.

I think it’s because Linux had no Licensing costs, and I’m pretty sure history bares this out. I don’t think Linux on the desktop contributed at all to its server popularity

https://www.wired.com/2016/08/linux-took-web-now-taking-worl...

BSD are staying relevant for its license (and for that zfs) and if it wants more adoption, it would be because of maturity on server usage and adding more manpower on desktop isn't.

I would use BSD if the server usage is more friendlier but right now, o don't see any benefit than using them for zfs.

Rewriting to rust and leaving the desktop space completely seems like a recipe for losing the bigger focus so resources can be spent on a fun project with questionable utility value for users.

> Although FreeBSD looks cleaner and more consistent than Linux, it's not really clear why should one use it over Linux.

Stability has a value all of its own. Back when I used Linux, every couple of years my system would get broken and I'd have to spend a couple of days fixing it. And often the commands for doing that would have changed since last time. Now I use FreeBSD and what I've got working stays working.

I had a love-hate relationship with Linux until I tried Fedora Silverblue. It's rock-solid for me and I can rollback on breaking updates with only a reboot or try other WMs with a rebase. Plus toolbox is great for keeping transient development stuff contained in a mutable docker-like (podman) VM.

As a longtime BSD user (since BSDI) I switched to Linux on my desktop several years ago, mainly for GPU support. It then didn’t seem worth it to use FreeBSD on the server and have to remember 2 sets of admin tools. So Linux it is. I think it could succeed with a usable desktop with accelerated graphics.

I’m (sort of) in your boat, but made a different decision - I’m a NetBSD user by (explicit) choice, so I choose my hardware to suit - which includes passing on latest hardware because of GPU support :/

> ... mainly for GPU support.

Is the nVidia driver still available for FreeBSD?

Many years ago I used to run the Linux binary of RtCW on FreeBSD using the emulation layer. Worked like a charm and the FPS was even slightly higher.

Yep. I am using the nvidia driver on FreeBSD 12.

In terms of non-nvidia, a few years ago the port of the Intel driver was badly out of date and a recent laptop would be stuck on vesa. But that issue has been fixed as far as I can tell.

> Honestly I don't think FreeBSD should focus on desktop story at all. Resources, as limited as they are, should be spent on making FreeBSD an obvious and compelling choice for the servers

This is how things bitrot and driver support fades.

> Honestly I don't think FreeBSD should focus on desktop story at all.

Lest we despair, macOS was based on FreeBSD. Also, lest we become overly enthusiastic, macOS is about the only successful desktop based on FreeBSD (apologies to y’all Dragonfly folks).

I wish there was more diversity in this regard but I think it’s an unfortunate instance of collateral damage of the GPL vs permissive licensing wars.

The MacOS userland tools were taken from FreeBSD, along with a bunch of other stuff, but saying it was 'based on FreeBSD' is a bit of a stretch.

NextStep pulled from Berkley BSD, so some of that lineage is still around. Darwin has pulled from both NetBSD and FreeBSD, themselves derivatives from the Berkley distribution.

A shared lineage is a very different thing, however.

Ancillary devices like the old Airport routers also used NetBSD

Until the last few releases, macOS also used IPFW, the firewall explicitly created for FreeBSD, so there was more FreeBSD in macOS than just userland tools.

Sure, and that's why I wrote 'along with a bunch of other stuff'.

Writing something like 'MacOS is based on FreeBSD' implies a situation more like Sony's OrbisOS, which isn't the case with MacOS.

> macOS was based on FreeBSD

Can you share some source of your statement?

> Can you share some source of your statement?

https://en.wikipedia.org/wiki/Darwin_(operating_system)

There's probably very little commonality left nowadays though. However, if you go back to the earlier days (Panther, Tiger, Leopard) and look at the release notes of Mac OS at the time, and the release notes of the contemporary FreeBSD version of the time, you'll see very similar text: I remember word-for-word paragraphs about things like NFS fixes in both.

Read literally anything about Darwin, NextStep, XNU etc... and you will see "FreeBSD + Mach = OSX"

Most of the command line tools in MacOS were ported from FreeBSD. I believe they took TCP from BSD too, but I'm not sure if that's from Mach.

macOS was based on NextStep. That said, there is a lot of shared code, and Apple is known to have provided code back to the FreeBSD project.

Sure, look up A/UX and Jordan Hubbard, Apple's former Director of Unix Technology.

It won't be that many years before I imagine we'll start seeing people asking to cite how the next forms of ChromeOS or Android were related to Linux.

There's no A/UX in Mac OS X.

Nor in NeXT, but it shows a history of Unix/FreeBSD development at Apple prior to the hire of Hubbard during the Mac OS X era. That is to say, his status as co-founder of FreeBSD and Apple hiring him wasn't merely a coincidence.

And to clarify my example, Fuchsia is set to be a Linux-less functional rebuild of Android, but it'd be a bit myopic to point out that "there's no Linux in Fuchsia". It's like saying there's no Steve Jobs in any Apple products created after his death. Less perhaps, but not none.

I agree. Linux is struggling on the desktop and BSD shouldn't spend manpower unnecessarily for desktop when its primary strength is on the servers.

But it probably makes the developers feel better by dogfooding their own OS instead of using another one but we have enough virtualization that one can run server OS on top of their most productive desktop OS.

Nah, FreeBSD lost server story in the cloud. There is a very little story in on-prem. There is way more risk choosing FreeBSD as your server OS than as your desktop OS.

This is coming from someone who use FreeBSD on daily laptop and weekend desktop. I would just not pick FreeBSD at work for anything other than my personal-work VM.

Rust is not an answer. There is zero kernel development support from both rust and freebsd side. I know that because I'm working on KPI for both as well as multiple userspace library wrappers.

FreeBSD has seemed to find a niche in storage appliances. A few big names run servers on the same instances (eg Netflix).

Reply on news.ycombinator.com