Hacker News Reader

Comment by linspace | Hacker News Reader

Seeing this and the other day about SerenityOS I have the feeling from checking the home page that focus is on certain graphics experience, and so I wonder why build a OS from scratch instead of a window manager over Linux, apart from the coolness factor. My question is what things can impact the user experience if you are allowed to start from a fresh design?

View on HN

Linux 20 years ago wasn’t what it is today. Low latency audio for example, that was something that BeOS in the late 90s could do out of the box, a similar experience on Linux required special kernel patches and some compromises or it simply wasn’t there at all.

Audio latencies of less than 10ms on consumer hardware were otherwise unheard of. Windows, out of the box, was still operating in the 100+ms range.

Yet after 20 years of Linux distros, we still see Linux struggling with getting a simple desktop together without X or Wayland crashing itself and logging you out because of someone wanting to tweak their desktop via a custom init script or systemd script.

Let me know when I can support 100% of all Linux distros and users like I can with macOS and Windows.

There's only one macOS and only one Windows. You can support only the big distros (basically Debian/Ubuntu or/and Fedora/openSUSE; for example that's what Chrome does) and let respective distro maintainers handle issues to other ones. Or just use Snap or/and Flatpak.

I would love to see a BeOS/Haiku userland on top, of the Linux Kernel. That would bring wide hardware support, including HW accelerated 3D (something BeOS was always lacking). It is not going to happen though, the attempts that were didn’t go far and porting the entirety of Haiku would be a gigantic effort that no one is volunteering for.

IIRC that was one of the two/three main ways forward discussed and planned for after Be Inc went belly up.

- OpenBeOS (Haiku) - rewrite from scratch

- Option 2 (can't remember if it had a name) - was just to Implement BeOS userland on a Linux, to get free access to drivers etc. Most of the Be community was not interested, and I cannot remember if they even got started..?

- (YellowTab - Some company had access to BeOS sources and started releasing and somewhat updating the OS on a commercial basis - but don't think they actually had the rights to do that....)

If only Apple had chosen to go for BeOS X over NeXT.

This would probably have meant that the Mac wouldn't have become so popular amongst developers. At least that was the reason for me to switch to Mac because it's a "proper" UNIX with a window system that actually works ("Linux on the desktop" has mostly caught up in the meantime, but 10 years ago that was quite different).

Haiku has a perfectly good POSIX layer; I expect it would have been fine.

BeOS was less compatible. There were also many improvements to it in the later versions. When NeXT bought Apple[0], BeOS was not even in its pre-release phase. Its heyday, such as it were, was really on Intel, with R5 and the Personal Edition, in the last few years of the 90's.

[0] For -$400M :)

Option 2 : BlueEyedOS / Cosmoe

Yes! That was it.

When on the subject of "things lost in the mists of time", did the original OpenBeOS lead (Michael?) ever return?

I would love to see a hardware company pick up Haiku and “do an Apple”: make a small set of focused computers with excellent support for a specified set of hardware.

Intel should do this. It's rounding error and they can use it to push the envelope. Intel should have done this with Linux, Solaris, Beos years ago to push Microsoft.

They won't.

Android, not joking, most of the original team are former BeOS engineers.

And it's still terrible at low-latency response compared w/ iOS...

That has been sorted out, but you need to make use of NDK APIs like AAudio, and naturally own a flagship device, so are the Google ways.

The NDK wouldn't be a problem, but requiring powerful devices speaks volumes about how lacking is Android in that field, when one can build low latency synthesizers and effects using a cheap old Raspberry PI 3. https://www.youtube.com/watch?v=YtU8jMtbplE

Yeah, but that isn't running a managed runtime on top, and all the security knobs turned on to mitigate classical C and C++ CVEs.

Most maker audio devices are making use of bare metal with Cortex and ESP32 chips, if the ultimate performance on the cheap is the goal.

True low-latency workloads on modern Linux (and 10ms is definitely low-latency!) still require a custom patchset (namely PREEMPT_RT). It's slowly making its way through the process of being upstreamed, but still.

I'm glad you mention audio because I'm going to spend 60€ on a dedicated microphone just because Linux cannot use simultaneously my Bluetooth headphones and their integrated mic without manually changing the codec (which additionally sucks when using the mic). Some problems improve but others simply never dissapear, like audio. Maybe there is a lack of interest or maybe is a design or a process problem, I don't know.

Linux plus the PREEMPT_RT patchset would make a nice low-latency kernel for multimedia and UX-focused workloads. It would still need quite a bit of user-level work compared to most distros, however.

Some other OS projects have legitimate reasons for starting from scratch, e.g. Redox being written in a memory-safe language. I suppose you could class Haiku's compatibility with original BeOS software as reason enough.

Not really, it's a social reason not a technical reason: those who are 'hardcore enough' to build a 'BeOS shell' like to 'own' their kernel too..

From an 'time to release' POV it'll take much longer, but it's not like they're expecting to earn money from the release..

Haiku is still here while all the BeOS clone projects on top of Linux/BSD kernel died..

#1 reason is probably nostalgia, and just the idea that you have a simpler, older-style OS. vs a modern OS just reskinned to look old. Personally i think speed: I see a 90s UI and I expect stuff to happen almost instantly on my modern Desktop. Also expecting less bugs because it’s a simpler OS. But in practice even though Linux is much larger and more complex it’s probably still faster and less buggy.

So probably just the idea.

EDIT: Should also add, probably because these OSs do some core things differently than Unix and Windows. Those OSs are constantly updated but the core features are still old, so the niche OSs could possibly implement them “better”.

> Also expecting less bugs because it’s a simpler OS.

That's not my experince on late 90's. You sneezed, and w98se BSOD'd.

You probably cannot add 1+1, that person meant the 90's style desktop means there is no unnecessary cruft in it, therefore it won't try to amaze the user with eyecandy, so operations will happen instantly. And less code means less chance for bugs.

> And less code means less chance for bugs.

The OpenBSD bug on Xenocara which allowed root needed more code, not less.

So this is the case with pledge(4) and unveil(4), you need (a little) more code in onder to sandbox setuff properly.

NT4 was much more complex than W98SE, yet NT4/w2k was much better on multitasking (by a huge margin). I've seen w98se crawl even with a Pentium 4 and 512MB of RAM with just 3 IE windows open back in the day, while 2k flied.

>The OpenBSD bug on Xenocara which allowed root needed more code, not less.

X was considered and still is a "hack".

>So this is the case with pledge(4) and unveil(4), you need (a little) more code in onder to sandbox setuff properly.

You don't write a graphical subsystem and then tries to sandbox it. Sadly nix advocates won't ever understand this logic.

>NT4 was much more complex than W98SE, yet NT4/w2k was much better on multitasking (by a huge margin). I've seen w98se crawl even with a Pentium 4 and 512MB of RAM with just 3 IE windows open back in the day, while 2k flied.

Let me compare apples with pears, i mean two different approach for OS development which have almost nothing common except the w32 stuff, and let me conclude the newer is better. I don't really see your point.

I woudn't say "newer" is better, but sometimes you need a bigger and more complex codebase to get more performance, stability and security.

And bugs.

Because we already have enough UNIX clones and a Window manager lacks the full stack development experience.

>I wonder why build a OS from scratch instead of a window manager over Linux

During the last 30 years it was proven tt is impossible to create a seamless window manager / desktop environment on nixes.

>what things can impact the user experience if you are allowed to start from a fresh design?

It lets forget everything about the harm, mess and pain the nixes made.

Mac OS X is proof that this can be done if you build a window system from scratch, without X Windows.

But I’m not personally on board the just-use-Linux train. There’s room for innovation at the kernel level.

Reply on news.ycombinator.com