Re: [MiNT] Shutdown() discussion

[Odd Skancke <ozk@atari.org>]

On Wed, 10 Dec 2003, Standa Opichal wrote:

> Hi!
>
> On Wed, 10 Dec 2003, =?us-ascii?Q?Joakim H=F6gberg?= wrote:
>
> >
> > >Hmmm, OS should support HW. As we see ARAnyM as a virtual machine that
> > >basicaly defines that the CPU instruction set is mc68040 compatible and
> > >nothing more we would like to get the OS support ARAnyM just like any
> > >other HW available.
> >
> > IMO this is the crossroad right there - many might feel it is all backwards to see MiNT,
> > our native OS, be adjusted to let an emulator/virtual machine use native features from the
> > (real) underlying OS that it is running on top of. Like Ozk pointed out, the original issue
> > might perhaps be a detail, but all things start in small.
> >
> > The principle is probably more important than the issue here. Personally, I think an emulator
> > should be responsible for behaving like the hardware it wants to resemble, rather than relying
> > on the OS to be changed to suite its needs.
>
>
> Well, you might have read the reply to your notes above the snippet you've
> quoted... So again:
>
> It is matter of the point of view. ARAnyM is not considered to be an
> emulator and therefore it is not supposed to implement each new HW that is
> available as well as it doesn't for ST, TT, F030 or whatever existing.

 No, this is not exactly what I meant either. Hardware dependant things
should go into external 'drivers' as much as possible. I think the best
solution (that would keep at least me happy) for the shutdownquestion is
to implement a generic kernel function, and have the actual hardware part
put into an XDD.

>
> Better, it might be considered (by those whom it makes sense) as an
> alternative "hard"ware, or virtual ware any thing you can imagine that is
> _not_ compatible with the "emulator" word meaning you are speaking about.

 Ok, lets play with this. If ARAnyM indeed is to be regarded as a new
hardware platform, try to put as much ofthe hardware-specific code into
loadable device drivers.

>
> The world reality? All major software service vendors are trying to move
> somewhere where it really doesn't matter what system you are talking to,
> but only the content matters there. Are we going to do chats about what
> should ARAnyM be from the OS point of view? In your opinion it should not
> exit. Fine.

 Not at the expense of having ARAnyM specific code inside the kernel.
There must be better ways.

>
> I don't want to care about people that don't have anything suitable to say
> in this case. If you all care about ARAnyM specific code not beeing
> present in FreeMiNT then ok, we can fork a CVS for that. We can put it
> there compile time optionally. But think... EasyMiNT would be larger due
> to a special kernel for ARAnyM. Installation procedures and system
> migration would become greater nightmare than it was in the KGMD times. I
> would say that I would not go into this again.

 Okie, if ARAnyM (or whatever new hardware) cannot be supported the
'correct' way (XDD's or perhaps we need an extended kernelmodule API
thingy), then please do fork the CVS.


-- 
 Regards,

 Odd Skancke - ozk.atari.org - http://assemsoft.atari.org