>>108477409 (OP)
cool. produce a hardware platform with publically available schematics, documentation and no obfuscation of the chips utilized by this hardware. maybe then.. sure. i'll draw the logo. but now? with today's slop and locked down hardware? christ no. waste of time. we need to start from scratch and with no interference from lawyers from israel and the united states.

>>108477424
The biggest problem is the GPU, it could start off with a CPU renderer until it's figured out.

>>108477409 (OP)
all the system sounds must be autistic screeching.

So, who's writing the drivers?

>>108477424
this

File: 1743747814452991.jpg (944 KB, 1920x1693)

944 KB JPG

Windows2000/NT6.2

>>108477409 (OP)
First: Are we going to implement age verification? Very important.

>>108477409 (OP)
What’s the governance structure? How many seats on the leader ship council? There should also be a project lead, and it should be a diverse female.

>>108477409 (OP)
We would need to establish a Code of Conduct that's welcoming to all people and those who don't identify as people.

>>108477409 (OP)
We need a better language first.

>>108477458
What a retard.

>>108477866
And more importantly: tabs or spaces?

>>108477466
>The biggest problem is the GPU
sad but true
>start off with a CPU renderer until it's figured out.
yes. and since we're not having to support the half century+ of unix nonsense we can write some blazing fast routines using 2026 CPU tech.

>>108478153
Tabs, but anyone who fucks up spacing will be permanently banned

>>108477409 (OP)
single threaded only, single process only. kernels and were a mistake

It refuses to boot on any indian subnet, first thing it does is on loading an IP stack check it's public IP address and shuts down and corrupts itself if it finds it is in India or Pakistan etc. Only supports ASCII.

>>108477926
based

closed source and obfuiscated before compilation only the single thread and single process can check the value in a register or it shuts down and corrupts itself. Supports raw sockets

>>108477944
Of course. Need to know where cunny hides.

File: 2026-03-29 09.16.59 rentr(...).png (85 KB, 797x559)

85 KB PNG

>>108477409 (OP)
Go for it. https://rentry.co/g7aofwhc

>>108478271
the sigle process must load and provide the equivanent of chromium, codecs, ms paint, a clock and an IP stack and wordpad and a spreadsheet, no printers, It has a copy of doom in memory at all times and uses the reguster values in memory for doom as hashes to encrypt and decrypt all values it's writing to other registers. There is no way to run other applications unless the entire OS is recompiled

All code must be spagetti code, no exceptions and synthal is limited to the C library. Extra appreciation to code containing gotos, comments are banned

Um im literally taking os right now in college and I already figured it out. At first I thought we could just use gpu cores and wouldn't have to worry about scheduling because your computer will just turn off if it ever tries to make more threads than cores but I guess you cant do that for some reason so I invented the best method, you have many small batch cores that uses sdlf, if there would be a miss the longest work remaining thread gets sent to the singular super beefy fifo massive clock speed core. We can use fpga to get the logic down or whatever before printing real chips.

>>108478391
single thread single process or go home, everything else is inrinsically insecure

File: HEGeDGebkAA_BRI.jpg (424 KB, 1673x2048)

424 KB JPG

>>108477409 (OP)
I'm making one, it's single process, ARM, no MMU and no compatibility with previous slop
I plan to make a cyberdeck using it

>>108478391
Oh yeah the individual cores on a gpu arent freaking addressable for some reason you can only send a batch of jobs to like a 100 for some reason.

File: 1749836844087429.png (592 KB, 618x968)

592 KB PNG

obligatory

>>108478437
Here's the link by the way
https://github.com/reshsix/vermillion

>>108478437
If you had 128gb of ram would fragmentation even be an issue? All your programs would have to know where they want to go in memory or would it be only one program allowed in memory at once?

>>108478495
Oh wait mmu just does paging you can still find the next section of memory big enough without it.

>>108478495
I have a malloc that I want to simplify on future to avoid fragmentation
Programs load on the next free space of the heap, which can be also occupied by "terminate and stay resident" type of programs (which I called libraries)
The plan is to use more static allocation than dynamic inside them, by providing good limits to things

>>108478445
Computers are either autism boxes, or data surrendering boxes

>>108478519
MMU point is just to protect memory zones and scramble addresses
I hate it because it doesn't let user programs interface with memory-mapped devices

Alternative idea: let's just use the Linux kernel, fork GTK2 and I will make a CSS stylesheet that makes it look more like Mac, but like, more Mac than Mac.

The guy who wrote UNIX did it on his own in a matter of weeks. Why does no one else do this anymore?

>>108479319

Come to think of it, forced cooperation, force teami-ification, corporatization, what have you, is kind of the problem. If you think about great works: books, art, software, the person who initially built it was usually a lone wolf. No interference, no arguments, no debate, just a singular, cohesive vision. There can't be a cohesive vision if there are multiple people involved.

So really when we talk about the "next UNIX" or the "next Windows", it really does have to be one brilliant guy lone wolfing something magical together and then it snowballing from there. Other developers come in as the project gains traction and the bones/architecture are done, providing a blueprint for the extensions.

>>108479300
Yeah, dude lets ignore that most problem are inherited from C and the UNIX mindset.

>>108479319
No wonder why it sucks.

>>108479351

There is like 0% chance that the best possible implementation of an OS happened right as computing was in an infant state and just starting to grow. It's just that we're scared and lazy and can't pursue anything else now, it's "'good enough" and got first mover advantage.

Why don't one of these infinity rich faggots like Musk hold a competition with prizes for whoever can write the best OS, figure out some kind of standard rubric in terms of performance, memory usage, reliability, and just let the mass of unemployed and underemployed autists go to town. Winner gets to spearhead his startup.

I'll make the logo

File: we-arrested.jpg (67 KB, 500x666)

67 KB JPG

>>108478067
>We need a better language first.
An all-lists (Scheme/Clojure inspired) Haskell (purely functional.)
>>108477458
What a clever.

>>108479444
>https://rentry.co/g7aofwhc
Or Erlang it is fucking fast designed for Ericson !

>>108477409 (OP)
If you were to build an OS for the next century, it would probably collapsed the thousand layers that have bubbled up into a hypervisor and loadable modules like wasm using a WIT system

>>108478305
Redox recently put capabilities into theirs. It might be the only real advancement in OS. While that article makes a lot of good points, the new semantics stuff is more prescriptive than real.

>>108479372
Screenshot this and tag him in it on Xitter. He could find this with his pocket change.

>>108479319
There's lots of OSdev projects like the one that was posted here, but they don't gain adoption because people want "le best thing", and to satisfy all the bogus requirements they want only windows/mac/linux can

>>108479372
Because nothing is the best at birth, an OS could win the competition then be shitty long term, which is basically what happens all the time with venture capitalism fund winners
The problem is wanting "the best", "the objective best" instead of "this one I like the most", because "the objective best" is sticking to the already tested and mature crap

first of all it should be based on superior runtime without retarded threading model.

after some studies of shared memory i conclude that new OS should have APIs to share already allocated memory. now all OSes have some retarded file-to-memory mapping which is also involves retarded page file concept. so

> no page file
> allocated memory sharing via specialized API (no file word in there)

>>108479319
>matter of weeks
because it was written for a computer than only had 4kb of ram. are you stupid or just pretending to be retarded?

>>108479516
And they are doing nothing wrong. I want a new OS for design elegance, but even an extremely well designed OS wouldn't have much better performance in any area except for very specific ones.

>>108479602
They are, once you optimize for sake of optimizing, you make a square into a circle, which is meaningless and forgettable
You don't need server level performance for user software, you just need the thing to work without lagging, which can be achieved by limiting features

>>108477409 (OP)
Linux.

File: plan9.png (45 KB, 560x315)

45 KB PNG

>>108478067
>What a retard.
Well if you insist on being non-Unix, you could look at Plan9 or BTRON for ideas I guess.

>>108477409 (OP)
>What would be the best architecture for this modern OS?
Just copy seL4, it's a nice modern microkernel, and a fast one at that.

>>108478596
Can't you just use mmap for that? Kernel driver exposes your device as a block device and you mmap to expose to your application or use block device API via ioctl, etc. directly. MMU also can virtualizing memory addresses so you can have codes all linked to same address and run them at the "same time". This also allows to share memory between codes e.g. two codes rely on same data linked at say 0x0000000008001000 then MMU puts that data there and both codes can use the same.

>>108479687
Problem is having to write a kernel driver when you just want to set 0x8000 to 1
You can use /dev/mem but the kernel may get in your way (need sudo, it can overwrite your memory, writing a memory address becomes writing to a file now), for embedded is a hassle

>>108477409 (OP)
Genera running on a Lisp machine.
http://lispm.de/genera-concepts

>>108478067
>We need a better language first.
There's Common Lisp, which could easily support languages like Python and JavaScript too. You could just share objects between languages.

>>108479701
C and Unix (which includes Plan 9) are why there is no reliability and security in software anymore. Buffer overflows went from something that can't happen in high-level languages and were only created by noob assembly programmers to being enshrined in an ISO standard because of C and null-terminated strings.

File: Nunally.png (705 KB, 845x633)

705 KB PNG

>>108477409 (OP)
Honestly it should be like a console.
Everything is read-only except one folder for dumping user files.
All programs are either already part of the system or "installed" by recompiling the OS onto a new single-write memory (as in, once written it becomes read-only).
Limit programs to 100 MB of data. There's no memory management except each program can be loaded or unloaded from a 100MB chunk of RAM.
Single thread, one process at a time except each process must implement the Exit routine bound to a standard key like Escape which returns control to OS. Memory corruption or other issue is fixed by restarting OS. OS loads in under 2 seconds btw.
Programming language? A simple wrapper over standard CPU instructions to give readable names and syntax. No "compilation", no "variables", no "scope", libraries are just code you write to memory (the whole 100MB is executable).
Logo? A digitized 30x30 binary pic of Arnold Schwarzenegger giving roman the salute.
Name? Something that scares normies away.

>>108477424
>uhh you can't do that goyim because t's really hard o algo
They're not sending their best

The best approach to making an OS is from the outside in. You build the UI and userspace inside a hosted OS, then you defer the boring stuff (drivers, syscalls, compilers, etc.) for the end when the OS is working.
Having said that, even if you make a good OS, there's the problem of having no software written for it.
Never mind the fact that /g/ would be utterly incapable of writing a good UI toolkit (actually you guys wouldn't even settle on a kernel architecture, you'd spend 5 years arguing over it before writing a single loc)

>>108480376
More like when projects here exist people are too busy imagining things instead of using/participating

File: 1772129431157361.jpg (102 KB, 540x574)

102 KB JPG

>>108480144
entertain us more, mr.Clown

>>108477466
Intel has public documentation for their GPUs

>>108477409 (OP)
We had some discussions about this in the past /aosg/ generals.

>>108480852
The Wiki is back online.
https://igwiki.lyci.de/wiki//aosg/_Alternative_Operating_System_General
https://igwiki.lyci.de/wiki/ShimOS

>>108478067
>We need a better language first.
Ha, the usual trap. I knew AI researchers in the 90s who said the same. The next 20 years they were building "better languages" instead of AI.

>>108479383
Damn it, i wanted to make the logo

>>108478445
>no kensington lock
disgusting

File: Plus.gif (1.6 MB, 500x500)

1.6 MB GIF

>>108477409 (OP)
>legacy bloat
Why not start out with modern Linux kernel and then rip out all the misbegotten APIs? Their API/ABI promise means it is accumulating cruft. Also the glibc is full of idiocy such as strfry. Rip it out.
SerenityOS has no API/ABI promise and can experiment as much as they want.

>>108481504
Idk what that is and I don't care (I'm employed)

>>108477409 (OP)
ReactOS is the way to go; Windows NT has proven itself to be a good OS in the 2000's, and there is so much existing software made for those old Windows systems.

File: VMS_Uptime.png (28 KB, 807x475)

28 KB PNG

>>108481762
Before that, VAX VMS proved itself as rock solid.

>>108481967
i bet this screenshot wasnt made from VAX VMS OS

>>108477409 (OP)
look at how CloverOS is going op

>>108477409 (OP)
Make it retarded simple.
It is very hard to make a mistake if there's only 500 lines of code.

>>108482359
and the only functionality should be the display of the logfag with /dmp/ trash playing

File: 1772227819115767.png (523 KB, 1200x1500)

523 KB PNG

>>108482374
That was uncalled for...

File: quote-unix-is-a-junk-os-d(...).jpg (51 KB, 850x400)

51 KB JPG

>>108477458
Counterpoint:

>>108483546
He never said that.

>>108483551
Cutler's hateboner for UNIX is well documented. There is a reason of why Windows NT is basically the Anti-UNIX OS.

>>108483558
And with good reason

>>108478251
No. It degrades enough to deter them, but not enough to get them to jump ship. The sweet spot would be if it drives them to an heroing.

ponyos and kolibrios both pique my interest but their build env is dog shit

>>108484079
KolibriOS is good, but it's only 32-bit and its development is quite slow.
MenuetOS (KolibriOS' ancestor) actually is 64-bit and has a faster and consistent development, but the creator of the project had a tantrum years ago and closed the source code. MenuetOS was originally licensed as GPL.