from DaemonForums

I don't need Bluetooth, games, any fancy code editors, Wi-Fi, printers, or gadgets.
At most, I want to buy a good home scanner.
In terms of interface, I only need USB for a digital camera and an action camera.
That's all. What's available in packages and ports is enough and even more than enough for me.
I don't need (as they wrote on the forum) a complex Linux. And Linux has indeed become complex.
FreeBSD is simple and more or less understandable. It's easy to fix. It's like a repairable system that doesn't need to be reinstalled
like everything else.
And I abandoned rolling releases back when I was still on Linux. So I don't need fancy and glitchy things.
Sound works well in *BSD.
 
FreeBSD is simple and more or less understandable. It's easy to fix. It's like a repairable system that doesn't need to be reinstalled
like everything else.

I don't think Linux is more or less complex than FreeBSD, although it depends a bit on the distro of course. There's differences for sure, and some bits are more complex, but that goes both ways.

systemd is more complex, sure, and that's usually the thing people point at. But you don't *need* to use systemd. And there's tons of non-systemd options.

In other ways FreeBSD is more complex than most Linux distros – have you ever looked at /usr/ports/Mk/? Those Makefiles are some of the more complex hardest to edit code I've seen to this day.

Some of the other complexities (elogind, devd, dbus, whatnot) exist on both.

I use Void not because I think it's necessarily the best Linux, but because I've been using it for 10+ years and can make it do anything I want. Back in the day I could do the same with FreeBSD (I even hacked those /usr/ports/Mk/ files). Often times what feels "complex" is largely a matter of familiarity.

Not here to convince you of anything – if FreeBSD works well for you then that's great! Just saying that it's not strictly more complex.
 
Linux (means, kernel only) and Linux "distros" are different things.
I don't think "Linux" to be too complexed, but Linux "distros" are not.

What I don't like in recent "Linux" are mostly KPI/KBI instabilities (if it's stable enouch, we should NOT need multiple graphics/drm-*-kmod and corresponding nvidia-drm-*-kmod[-devel] ports unless any old devices are dropped.

And another is instabilities in device naming strategy especially on network interfaces. When I last tried to configure Linux (in a distro), they were eth* only. But recently read on somewhere that significantly changed. What?!
 
At most, I want to buy a good home scanner.
For a scanner, I'll recommend checking first SANE: Supported Devices and then, if needs arise that you'll need near-pro quality scan cross examine with LaserSoft Imaging SilverScan Supported Scanners.

SilverScan is the best scanning software on market, it's amazing how much better results you can get from the same device with it, and it even comes with IT8 calibration/profiling if you need color accuracy (unfortunately, only works on Mac&Win)

Think in advance will you ever need to scan films/transparencies and will you need to scan bunch of printed documents at once, in which case scanner with a document feeder can be very handy.

As a rule of thumb, Epson models can be the best choice, but some of their cheaper (newer) models are not supported by SANE.

If you have any questions, please feel free to send me direct message here, I'll try to help as best as I can.

Good luck.
 
What I don't like in recent "Linux" are mostly KPI/KBI instabilities (if it's stable enouch, we should NOT need multiple graphics/drm-*-kmod and corresponding nvidia-drm-*-kmod[-devel] ports unless any old devices are dropped.

And another is instabilities in device naming strategy especially on network interfaces. When I last tried to configure Linux (in a distro), they were eth* only. But recently read on somewhere that significantly changed. What?!

Is that really so different from FreeBSD? Linux offers userland compatibility ("we don't break userland") and guarantees nothing when it comes to kernel interfaces. This seems like a reasonable trade-off, because if you start guaranteeing kernel interface stability you're just going to slow development down. And the userland compatibility is pretty good; last time I tried I could still run the last released xv binary from 1995. AFAIK FreeBSD doesn't guarantee anything on either kernel or userland compatibility between major releases?

The device naming seems like a superficial minor thing to me. You can endlessly bikeshed over the "best" way to handle this, but in the end it doesn't really matter IMO.
 
That guy made a pretty extreme horseshoe move. From OpenBSD to QubesOS and maximum containerization, or rather full VMs.

IMO OpenBSD, with its security-only focus, is pretty much niche. A switch to another niche O/S in QubeOS shouldn't be surprising.

As a baseline I don't think FreeBSD is any less secure than OpenBSD. OpenBSD achieves this by limiting default services. One would accomplish this with FreeBSD using the same strategy.

Much of what he wants be could done on FreeBSD with bhyve.

Or jails. Neither of which are supported by OpenBSD.

On a positive note, OpenBSD is where we get OpenSSH and other great software from. But an O/S is more than its individual software components.
 
I don't think Linux is more or less complex than FreeBSD, although it depends a bit on the distro of course. There's differences for sure, and some bits are more complex, but that goes both ways.

That's not entirely true. FreeBSD tries to adhere to the UNIX philosophy. And, FreeBSD tries to maintain the BSD directory structure. The various Linux distros are all over the map WRT O/S structure. That was my complaint with the Linux distros in 1995 and still is my complaint. There's the UNIX way and the Linux way.

systemd is more complex, sure, and that's usually the thing people point at. But you don't *need* to use systemd. And there's tons of non-systemd options.

This is certainly not the UNIX way. Sun Solaris with its services is a elegant approach that does not conflict with the UNIX way. Systemd is anything but UNIX. But then, the systemd author has advocated for the abandonment of UNIX compatibility.
 
That's not entirely true. FreeBSD tries to adhere to the UNIX philosophy. And, FreeBSD tries to maintain the BSD directory structure. The various Linux distros are all over the map WRT O/S structure. That was my complaint with the Linux distros in 1995 and still is my complaint. There's the UNIX way and the Linux way.


This is certainly not the UNIX way. Sun Solaris with its services is a elegant approach that does not conflict with the UNIX way. Systemd is anything but UNIX. But then, the systemd author has advocated for the abandonment of UNIX compatibility.

Directory structure also seems such a superficial thing to me that can be bikeshedded to infinity, but, in the end, barely matters. And a number of aspects here are just an accident of history as much as anything else. To be honest I'm not really sure what the "UNIX way" means exactly, but personally I think it's more useful to ask "is X a good idea?" rather than ask "does X fit some philosophy?"

I'm not a fan of systemd, but as I said you can use other systems such as runit or s6 (and there are distros based on it). You don't *need* to use systemd if you want to use Linux. And while I'm not a fan of systemd, it does work.

Look, I'm not here to defend Linux or attack FreeBSD; both systems are different and it's fine to prefer one over the other. All I'm saying is that Linux (in the broad sense, not in Linux-the-kernel sense) is not really more complex than FreeBSD, and that FreeBSD has its own complexities as well. Can you really hack those ports/Mk files? Or even rc.d, which has its own complexities (and, arguably, hacks)? For most mortals: probably not.
 
Recently, I was approached to work on some software for a Linux system. I also want to run some software that doesn't run on FreeBSD. I wanted to run i3wm on Linux but it's X11 only. The software I want to use is Wayland only. So I could use Sway but the Sway people don't support Nvidia drivers. Hyprland is too much eye candy for me and might not be up to snuff though I really don't know. Some of the software I want to use needs systemd. Some don't.

While some complain that FreeBSD doesn't run some things, Linux is a mess of incompatibilities.
 
Directory structure also seems such a superficial thing to me that can be bikeshedded to infinity, but, in the end, barely matters. And a number of aspects here are just an accident of history as much as anything else. To be honest I'm not really sure what the "UNIX way" means exactly, but personally I think it's more useful to ask "is X a good idea?" rather than ask "does X fit some philosophy?"

Sure. Let's propose we put FreeBSD in /system and our 32-bit compatibility files in /system32? I suppose that would be acceptable.

How about as some have suggested replacing /etc with a searchable sqlite database?

NO, there's the UNIX way and then there's not the UNIX way. When I first started using Linux 20-25 years ago my team and I had to remember that under Solaris, Tru64, DG-UX, HP/UX, Irix and the various BSDs the files were one place and under various distros of Linux the files were somewhere else. There's a reason POSIX was a thing.

I'm not a fan of systemd, but as I said you can use other systems such as runit or s6 (and there are distros based on it). You don't *need* to use systemd if you want to use Linux. And while I'm not a fan of systemd, it does work.

And a lot of people like systemd. Everyone I work with at $JOB loves systemd. They're not enamoured with runit or s6 and they certainly hate the Solaris and BSD approaches.

Look, I'm not here to defend Linux or attack FreeBSD; both systems are different and it's fine to prefer one over the other. All I'm saying is that Linux (in the broad sense, not in Linux-the-kernel sense) is not really more complex than FreeBSD, and that FreeBSD has its own complexities as well. Can you really hack those ports/Mk files? Or even rc.d, which has its own complexities (and, arguably, hacks)? For most mortals: probably not.

rc.d is an attempt to use a hybrid SYSV approach because the BSD approach didn't scale that well. Sun implemented svc to address weaknesses in the SYSV approach. While Linux went the systemd (everything including the kitchen skink) approach. Personally, the svc approach by Sun is probably the best approach. The reason we (FreeBSD) don't use it is CDDL. CDDL is ok for ZFS because if someone doesn't want to use CDDL they can use UFS instead. But add svc (to run underneath init) will mean we have to maintain rc.d and svc configurations in base. It's a duplication of work and that's a non-starter.

The various Linux distros do their thing which is not UNIX. People I work with have never used UNIX. They don't understand UNIX nor an old dog like me.
 
That guy made a pretty extreme horseshoe move. From OpenBSD to QubesOS and maximum containerization, or rather full VMs.

Much of what he wants be could done on FreeBSD with bhyve.
bhyve Type 2 Hypervisor vs Qubes OS (Xen) being Type 1. Perhaps an important distinction for people.

FreeBSD not being able to run OCI containers (no native Docker) seems like the largest problem this guy has with BSD-based solutions. He already had momentum to switch to something, so this decision makes a lot of sense. Surprised this has not been a higher priority to solve but perhaps not if people don't consider that advice such as "just use jails" comes across as a boil-the-ocean statement to the recipient, even if the person giving that advice doesn't see it that way.

Amusingly, several of this guy's complaints were desktop-oriented which people on this forum don't seem to value and complain when FreeBSD leadership tries to solve.
 
Back
Top