John has some very good points.
> >3) Once again, we were JOKING about that animation as:
> >a) producing animation for every bit
[ and other exaggerations about that animation]
On Thu, 15 Feb 2001, John Foust wrote:
> Sorry, I didn't catch the sarcasm.
Sorry. My lack of clarity. It can be hard to tell whether one is
exaggerating when finding fault with Windows.
> While we're on the topic of
> animation in OSes and apps, perhaps we should switch to the
> talking, dancing paper clip.
THANK YOU!! THANK YOU!!
You have identified _THE_ FPU killer app!
We shall name the paper clip "Float"!
Obviously, the PRIMARY reason for the FPU requirement is so that
"Float" can perform for us!
> I doubt this discussion has been heard within the halls of Microsoft
> for more than a decade, so I guess it's on-topic. I see your point,
> particularly from a classic standpoint... an OS is considered the
> part that runs apps and manages resources, not drawing pretty pictures.
I think that you're right about my attitude about the role of the OS.
> But it's like arguing that Linux isn't an OS, it's a kernel and a bunch
> of GNU apps wrapped in a distribution. Win 9x doesn't really exist
> without the bundle of everything else. Could it be written without
> floats? As an exercise. It would be more fun to argue that DOS and
> Win 9x aren't really OSes. :-)
Rather than "could it be rewritten float-free", I'm still aghast that any
float was ever put into it. (Am I too extreme?)
> I'm sure they didn't want to scare the bejesus out of the first
> wet-behind-the-ears CS grad who attempted to use "float" in their
> C program and got a link error. So where might floats be? In graphics
> code, particularly sloppy graphics code. But these days the OS includes
> OpenGL and DirectX, too, which have floats galore. In user-friendly
> code that showed percentages for completion or resources consumed.
Those percentages seem to always be displayed as integers. I teach lower
division undergraduates to avoid using floating point for coming up with
such numbers. Use of floating point (and then rounding to integer) for
calculation of percentage completion or resources is EXACTLY my point in
my bashing of Microsoft for incompetent programming.
> And then there's scads of apps like "Calc.exe", the calculator app.
Yes, there are and will be apps for which floating point will be useful or
necessary. I'm willing to live with slow floating point emulation to get
by for them.
I maintain that the OS could/should provide the ap with a "system
variable" for whether the resource (FPU) exists, but then it is the
responsibility of the ap to use/shut down/substitute as needed for the the
presence or absence of the resource.
What's next? The next version of 9x will refuse to install if there is no
sound card? Probably. "SOME apps might need it".
And yet, for all of their efforts to protect apps that might want float,
DEBUG (my favorite program from Microsoft) is STILL hard coded for 8086
ONLY code. DEBUG has been "upgraded" in numerous ways, such as additon of
XA, XD commands, but no provision for mnemonics nor code for instructions
that came after 8086, such as PUSHA, or SHL AX, 4
Rather than having to DB those instructions, I wouldn't mind having a
command line option or mode switching command to enable 386 and above
code. Or EVEN make 386 and above code the DEFAULT! (It could then be MY
responsibility to not use "new" commands when coding for old processors)
> >Any reason why I couldn't create a glacially slow machine running it on a
> >386SX?
> Uhm, it would be glacially slow? :-) And Intel / Microsoft cabal
> doesn't want you to even try.
Sometimes it feels like they don't want us to exist.
--
Grumpy Ol' Fred cisin_at_xenosoft.com
Received on Thu Feb 15 2001 - 10:55:55 GMT