What is memory protection and why is it so hard to implement for the AmigaOS?

[trekiej]

The Syllable OS team was creating a version of Linux for server purposes. It would have there own gui and libs, if I remember correctly.

[freeaks]

@trekiej
syllable isn't very amiga-like ..
what i was refering too was more like having amiga filesystem structure supported in the modified linux kernel, modified to have structure like:
sys:, c:, libs:, fonts: and so on .. instead of /usr, /etc, /bin which is the traditional unix structure..
at least .. and of course a GUI layer .. be it using xorg or something else but that should look and behave like amigaos:  workbench or dopus magellan lookalike
that should be possible to have something snappy and that look and behave like workebench using only xorg libx11.. on linux there's even a project that does this: amiwm .. with draggin screens too .. it's usable and very small and fast.
there's no file manager though, and many features are yet unimplemented, but as a proof of concept it's there.

it's too bad half the remaining amigans don't understand the MP necessity or what is it exactly, and the other remaining half seem to prefer solution like aros, os4 or mos which all replicate os3.x with all its weaknesses

where what i want is different. it's a system that's small, whose structure is logical and simple like amiga, where the gui is looking nice and customizable, snappy..
well ..an amiga-like os, minus the crashs ..





and about xamiga.net, it's just another emulation running on top of linux.
emulation is not good enough, there's still no MP. everything inside the emulated environement is just like usual. no MP, no resource tracking .. plain old os3.x 68k emuated.

be it by redesigning current kernels or using unix ones i don't  care but .. the crashs and instability i can't accept

anyway .. i just have to wait .. either all thoses system will eventually evolve or die.

it's just i want MP and resource tracking for yesterday..

[lsmart]

what i don't understand is why aros, os4 and mos users can't properly kill tasks and free their resources.

I believe the problem here is that ressources aren´t limited to the task claiming them. It is perfectly legal to have one task claiming e.g. RAM giving the pointer to another task, waiting for the task to finish and then releasing the memory. Or the second task may release the memory for the first one.
In fact there is a garbage collector for Amiga Oberon that does just that: tracking the memory of Oberon-programms and releasing it, when there is no reference left. Of course it will also free them if the task is killed. But that is because every allocation or copying of a pointer is tracked by a special library.

There is some beauty in both concepts.

[Piru]

what i don't understand is why aros, os4 and mos users can't properly kill tasks and free their resources.

Global shared address space. Everything depends on it, from the very basics of the messaging system. From time to time some people claim to have invented a way to introduce a reliable and bulletproof memory protection scheme to amigoid system. It always ends the same way, all the sudden these people fall ominously quiet and the idea is forgotten for the next couple of years.

If you change the very basic concept of the global shared address space you will completely break all the existing APIs and applications. You'd need to have completely new set of applications, which would most likely be ports of open source apps. If so, how will it be any better than YALD?

[freeaks]

Global shared address space. Everything depends on it, from the very basics of the messaging system. From time to time some people claim to have invented a way to introduce a reliable and bulletproof memory protection scheme to amigoid system. It always ends the same way, all the sudden these people fall ominously quiet and the idea is forgotten for the next couple of years.

If you change the very basic concept of the global shared address space you will completely break all the existing APIs and applications. You'd need to have completely new set of applications, which would most likely be ports of open source apps. If so, how will it be any better than YALD?

well why not ?
all linux kernel need is a good GUI .. see osx(i know mach kernel from bsd branch .. but it's unix familly still)
osx didn't reinvented the wheel, they just took kernel bits and bobs and made their sauce ..
and their product is kind of popular..

first thing first, such a modified kernel with amiga paths and filesystem structure would be a fresher from the traditional unix layout..
then a nice snappy customisable gui ... could be interesting ..

the problem most linux distro have is that their changes aren't very apparent, most run gnome or kde, and beside little changes here and there, for example what package manager do they use, or such small changes most distro are very similar to each other.
few like apple did real customisation on top of an unix kernel..

i'd bet something with the simple and logical amiga structure, and snappy customisable GUI ala workbench/magellan would interest many.
the merit are stability , and drivers ..
softwares like mui or reaction could be replicated to run on this new kernel, and become the basis for the whole interface, pretty much like cocoa does on macs, but done amiga style: small and efficent

well all that idea pretty much seems like what anubis would have  been.
but it'd not mandatory to use a linux kernel .. it's just convenient because it's there..
and "amigaoid" systems badly need a core update ..
backward compatibility it's good enough while transitionning .. see osx, at first they had classic environment running inside osx .. after a few year, classic env went away ..
problem with amiga, is that it frozen at that transition moment in time..
without completely achieving it
how long the wish for backward compatibility will hold us from progressing ?
68k software from 90s .. will it still be regarded as important in 2015 ?
i sure hope not ..
i hope dev from amigoid "ng" will make a clean table, and start making a solid base . where app and resources can at least be killed and freed.
i was reading about atari the other day, they have MiNt..
so even atari got MP .. unfortunately for them they got no new hardware. and probably an even smaller community. still they adressed this important problem quite a while ago. at first they were in the same boat as amigans..

[Piru]

well why not ?

Sure. Go ahead, I'll be waiting for your product release.

It's far easier to give advice than to actually produce something tangible.

And btw Apple used years on Mac OS X, and had budget of hundreds of millions.

[yakumo9275]

Lets also not forget all those years and millions before it has a name change and was NeXT OS.

freeaks, I'd love to test your

all linux kernel need is a good GUI .. see osx(i know mach kernel from bsd branch .. but it's unix familly still)
osx didn't reinvented the wheel, they just took kernel bits and bobs and made their sauce ..
and their product is kind of popular..

first thing first, such a modified kernel with amiga paths and filesystem structure would be a fresher from the traditional unix layout..
then a nice snappy customisable gui ... could be interesting ..

the problem most linux distro have is that their changes aren't very apparent, most run gnome or kde, and beside little changes here and there, for example what package manager do they use, or such small changes most distro are very similar to each other.
few like apple did real customisation on top of an unix kernel..

i'd bet something with the simple and logical amiga structure, and snappy customisable GUI ala workbench/magellan would interest many.
the merit are stability , and drivers ..
softwares like mui or reaction could be replicated to run on this new kernel, and become the basis for the whole interface, pretty much like cocoa does on macs, but done amiga style: small and efficent

well all that idea pretty much seems like what anubis would have been.
but it'd not mandatory to use a linux kernel .. it's just convenient because it's there..
and "amigaoid" systems badly need a core update ..
backward compatibility it's good enough while transitionning .. see osx, at first they had classic environment running inside osx .. after a few year, classic env went away ..
problem with amiga, is that it frozen at that transition moment in time..
without completely achieving it
how long the wish for backward compatibility will hold us from progressing ?
68k software from 90s .. will it still be regarded as important in 2015 ?
i sure hope not ..
i hope dev from amigoid "ng" will make a clean table, and start making a solid base . where app and resources can at least be killed and freed.
i was reading about atari the other day, they have MiNt..
so even atari got MP .. unfortunately for them they got no new hardware. and probably an even smaller community. still they adressed this important problem quite a while ago. at first they were in the same boat as amigans..

when you have completed it.

[Paulie85]

Wow, nearly 5300 views for this thread! Hot topic.

[psxphill]

osx didn't reinvented the wheel, they just took kernel bits and bobs and made their sauce ..
and their product is kind of popular..

OSX is a bad example. Apple needed a new OS and spent alot of money on it, without it they wouldn't have been able to jump to yet another CPU & would be struggling to compete.
 
Take a look at how they solved the problems:
 
Old apps are run in an emulator, we can use UAE for that.
New apps run in a completely different environment, this is where your problems start.
 
You could base your new environment on linux. To make it practical you'd have to drop any idea of keeping C: S: LIBS: etc.  An Amiga lookalike window manager is doable (in fact I think it's been done), unfortunately it looks pretty dated.
 
Or you could start from AROS, but you've just gone round in a circle. It's hard to do memory protection in AOS because the way it was designed makes it hard. Your only option is to change the API, but that one change will ripple through everything. Even if you ever succeed you'll end up with another AOS alike system that will get ignored.
 
Your only real option is to make AROS compatible with AOS4.

[yakumo9275]

Wow, nearly 5300 views for this thread! Hot topic.

and its posted/asked about once a month and still all these people just can't get over how easy it is to do and why hasn't anyone done it in the last XXXXXteen years?  Why all us programmers must be sitting on our ass thinking 'gee, that MP/SMP amiga OS topic is just too damn easy, I need something much harder to work on....'

[LoadWB]

Lets also not forget all those years and millions before it has a name change and was NeXT OS.

And before that it was a student project.