Re: subscribe fork-l

Robert Harley (Robert.Harley@inria.fr)
Wed, 22 Jan 1997 11:31:51 +0100 (MET)


>A more serious answer to what motivates me... making great software and
>making it available to a really huge market. I think the Microsoft bashers
>sometimes forget how much cheaper software is as the result of
>concentrating developer resources on a single system platform running on
>commodity hardware.

Great? Cheaper? I'm getting a new machine on which I'll run:

The best OS around in it's Linux incarnation: $0
The best C compiler around, gcc: $0
My favorite editor, emacs: $0
The best compression program, gzip: $0
The best ML compilers NJSML and OCaml: $0+$0
etc.etc.
------------------------------------------------------------
Total $0

(OK, in truth I may pay $25 for a CD with some of these on it just to
simplify matters).

The hardware I'm getting is a 500MHz Alpha clone for $6K to $8K
depending on the exact model i.e., 3 times the bang of a top PC for 3
times the bucks. For problems that parallelize well (and don't use big
integers or floating point or memory intensively) that's the same
value/money as Linux on an Intel box. However most of my problems use
big integers and don't parallelize well, and I sure amn't going to
wait 3 weeks to run a one-week job.

Single system platform? Right now I'm typing this in Emacs on... (let
me check, I'm at a humongous xterm not the console, I don't even know
where the console is)... a Sparc. I'm simultaneously editing a
program which I compile (with gcc) and run on Sparcs and Alphas, some
in this building, some 9600km away. When it's done, I'll run it on
those plus MIPS boxes, RS/6000s and PPros (under Unix). If I can get
time on an old Cray T3D, I'll use that too.

Why is a single system platform even relevant? Oh yeah, when you're
locked into proprietary non-portable systems you don't have much
choice. Good thing I amn't in that situation since all the software I
use is distributed as source and Digital Unix, SunOS, Solaris, Linux,
AIX etc are source-code interoperable and have free compilers.

Why does Microsoft not go in for open systems? I suppose that as long
as there are plenty of fools who buy the hype due to ignorance and pay
through the nose for crap software that crashes every few minutes, it
makes sense to take their money.

I occasionally use a PC with Windows 95 to telnet to some real
machines. About one time in three, the connection fails with "unable
to connect" without even dialing up the host. After doing that once
it keeps on doing it until I reboot the machine. When Windows does
manage to connect, one has the pleasure of discovering that it can't
even do terminal emulation properly. It tries to interpret many
keystrokes of the Ctrl-something variety instead of forwarding them to
the host. As if that wasn't bad enough, it automatically logs you out
if it thinks you have been idle for 20 minutes. That would be fine
except that frequently the "20 minutes idle" happens when you're in
the middle of typing. Apparently the measurement of time to within 3
or 4 orders of magnitude is a bit too complicated.

Does Microsoft deserve to be bashed? You be the judge.

>I actually discovered this list after browsing for information on
>Windows-On-Unix, to help substantiate or refute the (facetious) claim that
>in the future, folks will only run Unix so that they can run their favorite
>Windows-On-Unix emulator on top.

I have never seen anyone run a Windows-on-Unix emulator. It might be
quite a humourous experience.

Bye,
Rob.
.-. Robert.Harley@inria.fr .-.
/ \ .-. "Dances with bits" .-. / \
/ \ / \ .-. _ .-. / \ / \
/ \ / \ / \ / \ / \ / \ / \
/ \ / \ / `-' `-' \ / \ / \
\ / `-' `-' \ /
`-' Hit me with those laser beams. `-'