"What if" you have money to develop only one feature of AmigaOS?

Hi,

I would like to imagine an alternative to the AmigaOS system.
Can you tell me, if you had enough money to add only one feature to enhance AmigaOS, what would it be?

Please give me a few words:
I want "this feature" because "it can help us...".

Just don't say "multicore" as it is so obvious and is probably one thing we all want for better performance.

Remember. Only one feature. The most important for you.

Thanks.

List:

OS:

Memory Protection x5
Multicore x 2+
Posix x2
Virtualisation
Network support
Enhanced Workbench

Software:

Security (vault)
Web browser x2
Development IDE

POSIX compatibility

Full Memory protection

Full Virtualisering

And full Memory protection and full kernel protection

POSIX

Password vault with API like Keychain on the Mac.

In-built web browser which is fully up to date and comparable to Wayfarer. This to make it possible to use AmigaOS 4 for work, thus making it a daily driver alternative.

Better network support (TCP/IP stack)

Full memory protection

An extended user interface (Workbench GUI) We still only use one window without additional display options.

Interesting topic.

Like you said, there are obvious core OS things we would all
enjoy...

- memory protection
- SMP
- how about a native port to X86 or ARM?

But I really think the biggest bang for our collective bucks
would be to get ahead on the browser front.

Getting to where wayfarer is would be a start... Getting to
where Walkero was trying to go (before getting Weaseled)
would have been even better.

All in all, we're so dependent on the web these days, without
a better (if slow) browser situation, most everything else is
moot, IMHO.

It would be nice if it didn't have to involve rolling over for
goggles or getting a weasel oil bath.

Worried...

PJS

Memory protection
SMP-AMP
Standard Development IDE

@pjs

What do you mean weaseled?

Thanks.
I compiled them in the original post.

I'm happy to see I'm not the only one thinking about Posix. Without being a developer I'm pretty sure it would help in porting other software to our OS.
I just wonder how we could go posix without losing the Amiga low level way of doing things (threads, avoid the /dev/sda notations...)

@pjs

With enough money we would have any Chromium-based browser. No more Wayfarer, Odyssey or other browsers that are unable to keep up with web technologies.

Great thread. I have lots of thoughts on these topics:


POSIX: No reason this can't be done (except for true fork() I suppose), remember POSIX is not Linux and is a subset of what is usually found on Linx. Amiga has had 'POSIX' before in the days before POSIX or SUS was a real thing. Just in those days it was BSD ports, and standard C library for C89. The OS is already pretty closely aligned. GeekGadgets/ixemul is/was basically complete as of about 1992, it's just that after that people gave up and we couldn't keep up. And OS4's toolchain being gcc/binutils based means it is also pretty compatible with the GNU part of the Linux. VBCC's PosixLib and Olaf's clib2 have a pretty good stab at most of the C99 and extended API. We also need pthread, librt, psockets, curses or ncurses, iconv, libdl and all the user-unfriendly commands. the main issue of course is all the stuff commonly found in a linux baseline that is well above and beyond what GNU provides. It's a shame the guy behind AmiWin, MiamiSSL et al is a bit of a jerk and never open sourced his stuff. AmiCygnix however is very much a POSIX environment thats pretty complete for OS4. Shame its not better integrated / part of the OS itself really, although its better than WSL on Windows is.

Browser:
AmiBrowser proves there are possibilities here - run the browser engine on a linux slave and render the UI Amiga side. You could actually do this with x11, perhaps AmiCygnix could be made to show firefox running on a slaved rPi or through some kind of rabbithole interface if running inside qemu on a linux host. Endianness is the main issue. I'm surprised no one has tried it before.


SMP:
Everyone assumes we need and want SMP to run any thread on any core. But you could equally do what say, PS3 did, which is actually just have different cores running different 'machines' e.g. have one core running the 68k VM, one core running a chipset emulation, one core running say a PPC emulator allowing you to run AmigaOS4 classic or simply warpup for those classic fat binaries. The cores are so fast these days you'd be hard pushed to need more than one for anything other than NLE.

Memory protection:
We all know resource tracking and memory protection conflict with the amiga shared memory model. Amiga is a lot like the linux kernel in so far as to avoid context switches affecting performance everything is effectively running in supervisor all the time. The difference is linux manages that by tightly reviewing and testing all kernel code to a very high standard. Whereas any idiot can run what they like on Amiga.

But these features are also a crutch, there for when the software developers missed something, perhaps something obvious, perhaps a rare edge case. But still a crutch. And while modern Windows on ntkrnl is more stable, you still have to reboot windows servers once every 24 hours because of all the memory leaks, so it's not as if it lets you off the hook to write better software.

Personally I think the answer is to build these features into the C runtime. Indeed, the SAS/C libraries include most everything you need to do this. Developers generally didn't use them because either too hard or too much performance impact. The latter we don't have to worry about so much now as I'm sure very serious Amiga user is using at minimum a top end CPU, and more likely an FPGA or software emulation 100x faster than the original and with enormous amounts of RAM. So we can afford to trade a bit of raw speed and probably some memory efficiency for better reliability now.

I can think of ways a future Amiga could implement a memory safe version of its API and retain backwards compatibility though, using 'safe mode' libraries that are actually clients to a server based reimplementation of exec and friends. Something as simple as canary buffers on all system memory allocations for a start.

I say memory protection. I am not a full time developer or a skilled developer but I did some gui for my own use and other small stuff for my self. But it is very frustrating when you have to reset the amiga everytime you try a new exe that you just created because the entire system crashes... and trust me, you reset a lot when you are learning to code on the amiga..