Quote:

I don't agree.

1) simplicity
The underlying design of the OS would be more complex, but to the user, it will be as simple as before. If not simpler, as he will not have to worry about what hardware he is running the OS on.

2) speed
Of course emulating the virtual CPU is much slower than a running native code. But a JIT compiler has already proved to be more than fast enough for running normal not-so-CPU-intensive applications. UAE and Amithlon show what is possible, even when emulating the 68k, a CPU architecture not optimized for emulating.

For applications (games, compression, video, ...) requiring more power, compilation to native code (not JIT), will produce faster executables. I'm not sure how much is possible there though. There will probably be a performance hit compared to a true native application, but it's hard to say how much of a difference there will be.

Quote:

I'm not up to date on this. How does AROS address this? Why isn't the same possible for a VM?
It's always possible to define different versions of the virtual CPU, but this should obviously kept to a minimum.
Also, when choosing a 64 bit VCPU for example, instructions can always be executed on a 32 bits host CPU, using more cycles.

Salut

Quote:

All your suggestions would negate the simplictiy, speed, and charm. As it is now, you just write for the original WB3.1 plus a couple common extensions like AHI and CGX, compile for the system it'll run on, and you have a native ap that is small and fast. Writing for some theoretical machine that doesn't exist also limits you to the virtual machine. Say you make it 32 bits - are you going to have to redo everything when you later switch the VM to 64 bits? 128 bits? AROS already addresses that. Also, a VM means you are always emulating the code; emulators are never equal to the native speed.

Your idea's sound fantastic in principle but Amiga Inc tried that with DE. AOS 5+ was supposed to be the complete implementation of AmigaDE where as all software was platform independant. Computers, PDA, Phones etc.

The idea is possible (still)

And the philosophy is superb, but I don't think it will ever happen now, not in the short term. DE was a gem to program for (in my opinion) far more compatibility than Java etc.

I hope that one day TAO's dream comes true and there is total compatibility between hardware no matter what the underlying processor. But that dream is some years down the line me thinks.

How difficult would it be to port AROS to Intent? Is there anyone working on that at the moment?

Hello,

First I'd like to say hi to everyone. AROS-Exec is a great initiative and I hope I can join some interesting conversations on this forum :)

I'm not exactly happy with the current operating systems out there. Windows has improved a lot over the years, but I miss AmigaOS, it's simplicty, speed and charm. I've attempted to use Linux on a number of occasions, but except for command-line work, I don't see much improvements over Windows. KDE and Gnome, which you are basically forced to use, are slow and bloated.

The OpenBeOS project looks interesting but is still a long way from version 1.0.

AROS is another excellent candidate, except for acouple of major disadvantages: no memory protection or multi-user. But I see it is being considered, what more can I ask? :)

But personally, I think a new OS should basically run on any hardware platform. I've been thinking a bit about this.

Some points:
- applications should run on a suited virtual CPU (Java-like)
NOT a custom programming language, so that any programming language could be used.

- parts of, or complete applications could be compiled for the host CPU for increased performance (on installation for example), *from this run-anywhere binary*, avoiding all the potential problems and commercial disadvantages of having to compile from source code

- the OS itself can, for the most part be platform-independent, except for the kernel, or even only a part of the kernel - drivers for PCI devices can be platform-independent

- certain levels of the OS and applications can be defined:
* 0 watch
* 1 cell-phone
* 2 PDA
* ...
* n PC

an application designed for level x can be run on any OS implementation level > x. A simple PDA game (level 2) could run on a PC for example, but a large PC game (level n) can not run on a PDA.


Mmm, that's all I can think of for now :)

I know that the implementation is not trivial, and most likely not even an option for AROS 1.0 anymore. But I think it's something that can be an important feature in the future. It'd be a interesting topic to discuss, at the least

Regards