Decreasing compatiblitiy of OS4 a good thing?

@OlafS25

I’m in between.

I see no point in dropping support for legacy programs. But I do see the need for change.
if look at the belly of the beast it's not pretty. From security and system stability point of view.

This is about system stability... the OS simply does not care about concurrency well, it was never built for SMP.. asking developers to manage L1 cash manually, in all shared list and public structures as well as message structures. Is a big ask, and not likely that is going to be success.

We might even want a true 64bit OS, the web browsers need all the memory it can get.

Compatibility can be maintained by sandboxing, it’s not same as emulation, we not talking about CPU emulation or hardware emulation, we just translate API A to API B.. if possible. Basically, a window manager that talks to host OS, and handles IO, keyboard, mouse, serial, parallel, network.

The main reason why it’s necessary is system stability, it’s far too easy to crash the OS, corrupt memory and file systems.

I’m not too worried about hacking, we have lived with this OS for decades knowing it limitations, and community has been tiny, but changing how the process stack works, can improve security. And improve stability. I think it might be worth it, even if you break a few eggs.

When it comes to AmigaOS3.x I see no reason why we can’t bring it on for the ride.. encourage better API’s there ensure portability, I see no reasons why improvements in OS4, cannot benefit OS3 as well.

Becoming UNIX / Linux… no that’s not the point… nor is it a good idea to build a card house on top..

@LiveForIt

Ditching 68k has advantages and disadvantages

To name a example where I come from... AROS

it is about being API compatible but not binary compatibility. That finally means integrating 68k only by emulator. The advantage you can basically make changes to the OS without caring about breaking compatibility with 68k. That makes it possible to do currently a migration from 32bit to 64bit. But this of course comes with a price. You can as example not use original 68k libraries in the system like those needed for arexx. I think this is possible in amigaos because of petunia. There is a replacement for it but it is not fully compatible. You also must run emulation to use 68k. That is creating a different feeling at many users obviously.

So short... if you want to move forward you have to ditch 68k compatibility and the way 68k is currently integrated in but it will also feel different if you still use 68k apps.

I do not know if 68k integration like it is currently done in amigaos has still the same meaning as it had in the past. I know that at least in past 68k integration was a reason why people preferred one system to the other.

After I wrote the message below, I realised everything has been said already.
I'll post it anyway.

I am using a lot of 68k software too.

WordWorth
TurboPrint
ArtEffect
AREXX
CubicIDE
PageStream4 (Since PGS5 is buggy at least)
To name a few.

What bothers me most is being stuck with a system where a single misbehaving program can bring the whole thing down. Then there's the nice hardware with multiple cores and being limited to just one.

The OS should move on.
Then developers would also be more interested in it when better development tools are available, and this too seem to be related to the OS being stuck too much in the past.

AmigaOS 3 is being developed further, so there is the alternative with Vampires and other "68k" hardware.

All other OSes have moved forward; we should too. To me, it seems silly to have two parallel OS 3.x and 4.x, with the only difference that one does run natively on hardware that can only be supported partially, at half the speed.

To have fairly modern browsers, especially since a few of us in the thread want one, we need to move forward. And we need a 64bit os for modern browsers. I don't think there is a way around it. we might be able to to get along for limited amount of time, but sooner or later nobody will be able to to take webkit and make it run on 32bit.

I vote for moving forward, but I wish a better and faster e-uae was available.
Judging from the explanations of kas1e, I think that is a good way forward. So count me in with kas1e and Andrea. and thanks to LiveForIt for cooking up some mystery better development tools.

@OlafS25

Quote:

it can be quite fast... and feel quite normal, if you don't go for a full emulator. A600GS / A1200GS - AmiBench is moving in that direction.

@LiveForIt

Amibench is as I understand it a Aros 68k distribution, similar to apolloos or my distribution, running in full emulation. The problem is not how fast it is but how good it integrates and how the feeling for the users is. Is it still feeling very much the same or not? Feelings are very important to retro users who we all are today in one way or the other.

@OlafS25

a mix of different things.. System54 as AmigaKit calls it.

but the point is it depends on ARM based libraries
and acceleration from the host OS.

@LiveForIt

Yes it is

But how is it related to the discussion if it makes sense to drop compatiblity to 68k meaning 68k software would no longer work outside emulation

I 100% agree with Kas1e here. If OS4 doesn't do something then it's a dead duck that's just never going to progress. And since I do understand the sandboxing concept, even if not in detail, I'm all for it.

I've just sold my OS4 system. And I won't lie, one reason, but by no means all is that OS4 is just flaky as hell.

I'm not a long term user. I only had A1200 briefly back in the 90's when they came out and have only been into OS4 for about 12 years now, so I don't necessarily know all the gory details of what I should and should not do and trust me when I say, it's extremely easy for me to crash the system. Especially when I start to mess about with programming.

I'll be back to OS4 at some point, I'm sure, especially if I start to see movement in releases and improvements so anything that can move things forward is a plus in my eyes.

I thank all the participants in this discussion. As a user who is not fully aware of what goes on under the hood, I appreciate that there could be ways to make the system better. I just don't want to loose the ability to run system friendly 68K programs.

I disabled Kickstart/petunia.library.kmod in my Kicklayout in June 2020 and I have been running my OS4 system without petunia ever since then (5 years and 8 months).

;MODULE Kickstart/petunia.library.kmod

Even with petunia disabled, some magic in OS4 allows me to run the following 68K software:

BackUp software by Daniel Westerberg (onyxsoft)
Final Writer 97
TVPaint
PPaint
Cinema4D-Demo
ARexx
RxMUI
rexxbgui.library
rexxtricks.library
MicroRexx
AWNPipe
SimpleMail
YAM
Newscoaster
AmigaWriterDemo
PostIt (note program)

There are some others but these are the most obvious ones.

I use ARexx and RxMUI daily in my scripts and ARexx programs.

BTW, AmigaOS 4.1 FE Update 3 is very stable on my MicroA1.

---
redfox

@redfox

As far as I know, Petunia "only" does the JIT. It means it speeds up the 68k emulation. Without Petunia the 68k is emulated slower.

I couldn't see Petunia being the culprit of instability on my system. Instead of disabling it entirely, you could add the apps you wish to
to the "Compatibly" Preference in the "JIT" Tab. These programmes are blacklisted and won't use Petunia, but will rely on the slower interpretative emulation.

@Tuvok

If you want to modernize it is not about turning on / off petunia. "Modern" in todays terms mean 64bit, SMP and MP. And you have to change the API itself. All means 68k integration in current form no longer work and have to be added by emulation. As I understand it currently software that runs system compatible can be installed like on 3.1 and works. That will be no longer possible. 64bit also would mean that you need a kind of 32bit runtime for older apps. As long you stay on PPC. If you make a ISA change emulating PPC will be more challenging.

@MartinW
Quote:
I hope the new owner has some more luck and stability with it in his/her life and I wish you good times with your guitar purchase.

Quote:
Now I'm a user of an X5k since it was first released, heck even my name was on the silkscreen, and I have ALL versions (4) of the OS installed and active. In my experience stability has improved in leaps with every release. And that while I do quite some coding and experimenting with it. So the term 'flaky' comes not in my vocabulary when it is about OS4.1.

I hope that both you, while having the system sold in favour of a guitar purchase and the new owner of your system have good times ahead. Hope to see you again here as an owner of another Amiga system.

OldFart

@Tuvok
E-UAE today has JIT as well, which is written by the same author, Álmos Rajna, who wrote Petunia. It is even better than Petunia (he added better and more modern tricks in the JIT code compared to how it was when he wrote Petunia). I know this because I talked with him during the process and helped, together with others, to beta test JIT for E-UAE.

So now, from the user’s point of view, all that is important is that when you run a 68k binary, it just has its own window/screen/whatever, just like a real OS4 app. There is no problem whether it is running through Petunia or through E-UAE; both of them have JIT today, and in UAE it is even better. All that is needed is to adapt running things through E-UAE a bit, so it will not show an “E-UAE” window, but will feel the same as it does now, running like a native OS4 app (visually, of course).

So users will have the same experience as now, but the OS itself can finally get rid of things that make it harder to maintain and expand for developers.

It’s not that I want any SMP or 64-bit support—that would not make a big difference anyway today—but at least by ditching the 68k parts from the OS internals in this way, it will simply clean things up and (maybe) prepare for some further expansion

@MartinW
Quote:

Now when Ben is kicked off, things seem to be organized and progressing better, and I as a beta tester can say that a lot of updates have happened lately, but, of course, OS4 will always be in a split state: some OS parts are owned by Trevor (kernel, 3D drivers, Warp3DNova, OGLES), and all the remaining ones by Hyperion, and I don't see how one or the other will buy things from the other. So that all makes the release of updates, etc., long overdue. But at least I can see that pretty much progress has happened lately, especially Trixie doing much work on Reaction and co. lately as well (the area many of us have always wanted to improve).

So things will never be easy (and they never were), but at least it is not dead still, that’s for sure. Taking into account that time tells, and many Amiga users have simply passed away and that number will only grow, but we are still here and the OS is still progressing in some way to give some of us a bit of satisfaction.

One thing I should have been clearer about. I'm aware that things progress, but that stuff just doesn't always get out to the users. It's almost like the beta testers are in a different world to the rest of us. At least on a different OS.

Now granted, I know how beta testing works, I'm a development manager after all, but the gap shouldn't be anything like as big of a void as it is with OS4.

Anyway, that's OT I guess.

If I understand you correctly, you want to separate/sandbox legacy 68k code more than it is done with Petunia now. So, instead that 68k code is directly accessing the system, it should access an emulation layer in the future.

I guess this will break more legacy 68k programs. I cannot imagine that Vinced or tools like GoSnap will work anymore in such an environment.

Just curious, all talk about this topic, where should OS4 go in the future like OS4 is still developed and is on schedule and rocking (sorry, could not resist). What have I missed, what is the roadmap, is there a team working on OS4? When will next version be available, are we talking.... two more weeks? (sorry again)

@xe1huku

Yes, development of AmigaOS 4.1 has been halted

You must have missed that Hyperion released Update 3 at the end of 2025.

https://www.amiga-news.de/en/news/AN-2025-10-00087-EN.html

Of course, further development is uncertain, as already explained by
@kas1e. The most important components of Aos4.1 include AEON (kernel, 3D drivers, Warp3DNova, OGLES), and that is precisely what makes it almost impossible to further develop Aos4.1 in these areas.

AEON does not deal with the x1000/5000 hardware in the form of firmware updates, etc. Unfortunately, the situation is as it is, and I understand that some users are very upset about this.

Maybe a cold shower.
PPC is a dead platform. Its limitations have classified it as retro hardware.
There will be no revolution here, only a “dignified death.” This is hardware for enthusiasts.
The history of AOS4 is full support for “68k” programs.
Let this system remain as it was and as we remember it.
All other improvements are welcome.