Re: [MiNT] Shutdown() discussion - Conclusion

[Lonny Pursell <gfabasic@hotmail.com>]

on 12/13/03 5:51 AM, Frank Naumann at fnaumann@cs.uni-magdeburg.de wrote:

> Additionally there is also the idea to provide special high level features
> like jpeg decoding and such things. This is a great idea too. To reach
> this for user applications the operating system must provide these service
> so that the applications can use this, for example introducing a /dev/jpeg
> for jpeg encoding/decoding. These device can now be implemented as DSP
> driver on Falcons or use a NatFeat on the emulator. As the device driver
> is part of kernel it don't bypass the kernel and all security and
> permission checks go the correct way. Just as an example. This was the
> main idea of the aranym developers and I'm sure they agree that all
> services must be integrated into the operating system for consistency (the
> point where the hot discussion started I think).

I pretty much got that impression.  My question as was asked before to Petr
that went unanswered, so I'll ask it again from the end user view point.
That being someone who uses only real Atari based hardware.  Assuming this
is implemented somehow, where will a natfeat call end up on a real machine?
I was also told the old machine could call natfeats to.  Is someone going to
provide this example jpeg decoding in some module form for a real machine?
If that is the case does it it not duplicate the same functionality as a
sharedlib?

Or will all such cool advanced features be deemed to slow to bother with
on real machines?  Thus it won't ever be used on real machines in the end?
Or is it aimed at the "what if" new hardware appears?  Which seems unlikely.

What I figure will happen is, I will see the kernel grow, not necessarily
bloat, maybe it can be done in a small way, but ultimately I really suspect
someone with real hardware will not see any new advanced stuff they can
call.  The natfeat stuff is mostly aimed at shoving stuff through the
kernel to talk to the otherside (host machine).

I am not automatically against new things.  I am against things that
won't benefit the entire user base of real machines.  If someone wants
to explain in detail to me how a TT for example might actually gain from
this please jump in.  So far I have not been convinced old machines gain
anything but a bigger binary.

-- 
Lonny Pursell    http://www.bright.net/~gfabasic/