OSDev.org

This thread is for members to describe - if possible - what they see as the ideal for an operating system, both regarding the technical aspects and the user experience.

Please avoid harsh criticism of others' posted opinions - no two members are likely to agree on these matters - though asking about details and politely offering caveats is fair.

64-bit ToaruOS

Lightweight installation
Super modern looking consistent flat UI/UX (like mac OS)
Without spyware
No libraries, only one file to install a program required
Scales well across multiple architectures
Has the same performance/responsiveness on all PCs, meaning no unnecessary code that would slow things down
Simple and intuitive to use
Doesn't crash
Uses C, C++ and Assembly for almost everything
Drivers are easy to install or even better are automagically installed

yup, dreams are dreams because they aren't real

I'm trying for yet another POSIX compatible OS. But that's mostly because those who don't follow POSIX are doomed to reinvent it, and usually badly.

Mostly I'm wondering how many things I can remove from Linux without sacrificing functionality. To that end, my OS will not feature virtual terminals. There'll be a frame buffer, an input device and pseudo-terminals, so if you want VTs, you can create them yourself. However, this all means I don't need a font renderer in the OS kernel itself. And for debug, I have the serial line, so that's a plus.

There will be no multi-lib, at least not for now. I.e. you're going to use AMD64 and like it! Thought I'd make a companion OS for PPC32, though, as that seems like an interesting architecture. And more sane than x86. But I have to get my hands on the hardware first.

User experience will probably be very similar to Linux, as that is what I know.

@Octacone: You are describing Linux's aspirations there. And, I mean, it is user-friendly. It is just picky about its friends.

Windows 10, but in an alternate universe where I don't have to consider doing a full image of my boot SSD before I do a release upgrade. I have yet to have any major or even minor-but-problematic issues with this OS and it blows my mind how many claims there are that it deletes your 70s North Korean gonzo porn collection if you sneeze with one hand behind your back on the day before a full moon.


More to the point, though, what is a dream OS? An OS that does everything you need it to? Because I need an OS that is stable, straightforward to use, and plays Windows games from this century while utilizing my 8700K and 1080 Ti as best as it can, but on another machine I need an OS that can route multiple gigabits of IPv4 traffic and NAT out a gigabit interface at line rate, while simultaneously being able to be hardened and managed over an SSH connection or a console line if things go really sideways. I don't see any way to have an operating system that does both of those things, and does them both equally well (and not in the sense that "doing both things terribly is still doing them equally well").

If I didn't play games on my PC I would probably run OpenBSD as my one and only operating system, because it fills all the boxes I need *except* push 144 Hz on a 1440p screen in modern Windows games.

Boot up instantly (i.e. less than a second). Then let me do whatever I want to do, without having to install any 3rd party software or download any packages. Then when I am done, press the power button and shutdown immediately, or let me unplug the machine and nothing is lost or corrupted.

Oh yeah, I want it to work exactly the same 20 years from now, no matter what I do to it.

Essentially, I want a C64, but with modern hardware.

Is that too much to ask?

There is no such thing as a "dream", or ideal, OS. What works for a tablet of phone is not what I would want on a desktop. What suits a server may not make for a good development environment.

Is there a more mundane dream to have than about the tools we use?

Takes 4K bytes ( 1 disc sector ) , capable of running all my favorite programs fine. vim, stronghold hd, age of empires 2 etc

Hi,
On a more serious note, modern systems ( even embedded systems ) are layered and modular so that it is possible for large group of people work on them and can be easily managed. I was wondering if a case study has been done where we could remove the extra layers and get serious improvement in performance. I am simply not talking about layers in operating system, but the system as a whole. Perhaps exo-kernel design presents a case study. I was thinking if there is a simple abstraction that fits neatly for everything. I find the forth model impressive- Everything operates on the stack, but something similar for system design.

--Thomas

More seriously, given your comments not just about Android (which I have no intention of defending, TBH, but I am guessing my beefs with it are not the same as yours), but about a number of other systems as well, I am curious as to what you consider a good operating system, or perhaps, what you would want an OS to be like if no current ones are acceptable to you.

I do see the reference to an NT-like OS in your .sig, but I'm not sure if that is really indicative of your real goals.

This probably deserves a separate thread, where we can all try to explain what we want most in an OS; in fact, I'll make one right now, and invite you and everyone else to it.

As a user, I like XP most. I am not demanding and pretty ascetic, so I strongly dislike everything bloated in size and especially - in CPU usage, no matter how many times it is open source.

Speaking about "dream OS" from the osdev perspective, then yes, my signature is indicative enough, at least as much as I am able to do that. you don't wonder, that almost everybody here and their dogs/cats are tryna create yet another NUXI/linux clone, but you wonder about my inclination to NT? Minimalstic and efficient NT-like OS for "wierd" CPU architectures (they are not weird, they are uncommon for osdev, since everybody here wants to conquer x86 desktop with their linux clones) - my dream.

Oh I could write reams on this subject. Tomes, perhaps. I'm just going to write a little bit, using this quote as a springboard:

nullplan wrote:I'm trying for yet another POSIX compatible OS. But that's mostly because those who don't follow POSIX are doomed to reinvent it, and usually badly.

My first reaction to that was, "You should try Plan 9!" But then I remembered the statement is true of an awful lot of systems. Still, some are better in some ways. In certain areas, Plan 9 achieves more flexibility and user-accessible power with vastly less code.

My dream... is to one-up Plan 9 in the same way! Am I insane? It's certainly a difficult research project.

nullplan wrote:@Octacone: You are describing Linux's aspirations there. And, I mean, it is user-friendly. It is just picky about its friends.

I love this line!

Octacone wrote:yup, dreams are dreams because they aren't real

I think they are not far fetched as you might think. For example OS/2 and BeOS fulfill all your dreams. They were shut down for that reason. (I'd add AmigaOS too, if it weren't for the spyware thing.)

Let's examine your points one by one:

Octacone wrote:Lightweight installation

Most certainly doable. I like Haiku installer for example. Simple, small, works. Debian has two flavours of installers: full and netinstall. The latter is very small.

Octacone wrote:Super modern looking consistent flat UI/UX (like mac OS)

Not a problem either. Only a matter of having some designer talents and/or a good talented designer friend.

Octacone wrote:Without spyware

Certainly can be done as long as no big money corp sponsors you You'll need a lot of hacking experience and decent IT security knowledge too, so it's not easy, but doable. Limiting which app can access which resource (the network for example) helps a lot, like in SE Linux or Solaris.

Octacone wrote:No libraries, only one file to install a program required

This one seems strange. Why? You could statically link everything, but that contradicts the lightweight installation criteria. You shouldn't be afraid of shared libraries, all you need is a decent versioning solution, like the one in Gentoo for example. Or force strict linkage rules on the apps like MacOSX and Solaris does.

Octacone wrote:Scales well across multiple architectures

Absolutely doable, only a question of algorithms you use. Choose the right ones, and the OS will scale well, no matter the architecture. BSDs and Linux are good examples (mostly, but not always, Linux is also the best worst example in many cases, think of BKL for example).

Octacone wrote:Has the same performance/responsiveness on all PCs, meaning no unnecessary code that would slow things down

Same as above, only a matter of choosing the right algorithms and abstraction layers. Choose wisely and your OS will be responsive no matter the architecture.

Octacone wrote:Simple and intuitive to use

This cannot be done. People are subjective, what is trivial for one of us, is complicated for another person. You can aim at simplicity for sure, but there's no ultimate solution here. You'll always need man pages or some other kind of help system.

Octacone wrote:Doesn't crash

Requires big experience, lots and lots of tests, but doable, no magic involved.

Octacone wrote:Uses C, C++ and Assembly for almost everything

I see absolutely no probs with these. Maybe I would add a script language to the list, that's all (Lua for example).

Octacone wrote:Drivers are easy to install or even better are automagically installed

Certainly doable. BSDs and Linux kernel modules are easy to install, and they can be autodetected too. I had no driver problem with Linux in the last 10 years. Having many drivers for your kernel is the issue here, not their installation, imho.

I think we all know what we want from an OS. That's not a dream, the only problem here is user's expectations and big money corp interest are not the same.

Cheers,
bzt

bzt wrote:

Octacone wrote:No libraries, only one file to install a program required

This one seems strange. Why? You could statically link everything, but that contradicts the lightweight installation criteria.

This reminds me of my past. I used to think I wanted lightweight software, thinking disk space was an important part of that, but eventually I had to recognize that I really wanted software which was "light" in terms of time and intrusiveness. Disk and memory space was a much lower priority, and trying to minimize them often made software worse in more important ways. I think most people who use the term "lightweight" haven't realized the distinction.

Static linking was the touchstone which made me consider all this. I concluded I'm not opposed to static linking. It makes program execution much simpler and faster. Static linking does make GUI programs reveal their true sauropodian scale, showing this is the Jurassic era of software -- unique skeletal structures to accomodate their immense weight. When I was still thinking of developing in C for Plan 9 I wanted to write servers rather than libraries for all this GUI bloat, but I'm not sure I consider it worthwhile any more. The words "bloat" and "lightweight" are like "lag", they mean so many different things, it's not very helpful to use them. But this is a thread about dreams, and if you can't be vague in dreams, they're not dreams.

It would be a mistake to think that the only, or even most important, function of dynamic libraries is to save disk space or memory footprint. Equally important is the ability to modify library functions - correct bugs or improve implementation - without having to recompile programs that use those functions.