RE: Freedom from Microsoft != Freedom from Intel.

I Find Karma (
Sun, 22 Jun 97 12:51:49 PDT

Notice the subject line. It's not Microsoft alone who is the culprit in
our many-crashed machines. Part of the problem is that these Intel
boxes --- "state-of-the-art dual processor 200MHz PPros" --- shipped
with these cruddy $3000 graphics chips that are allegedly optimized to
OpenGL but suck for everything else. Poor NT 4.0 didn't know what hit
it; it tried its hardest to deal with this newbie Intel architecture,
but there's only so much you can do in software, right?

> We have like 30 machines in our lab, and they DO NOT CRASH.

My Win95 toy at home doesn't crash. Well, it does crash, but only once
every few days. But I don't use it for anything other than PointCast,
an occasional Xterminal, and training Michelle.

> Except we had one machine that would crash every couple of days. I
> finally narrowed it down to a memory problem, replaced the memory, and
> everything was great. For a couple of weeks, at least. At that point I
> realized it must be the power supply frying things, in particular the
> memory. But Jim refused to let me fix the machine any more, so back to
> Redmond it went. Three months later I found out that the machine was
> given to a new hardware support guy as a puzzle, and he finally did in
> fact track it to the power supply, which he replaced, and the machine
> was fine after that.

So maybe our problem is memory. Maybe it's the bus. Or the cache. Or
the graphics card. Or the motherboard. Who knows?

I know JoeK's NetBSD machine has only crashed once in its tenure, and
that was due to Quake.

But in the 2 months I've had NT, my OS must have "locked up" (that is,
ctrl-alt-delete won't do a thing) 50 times. I've had this weird
capslock bug that comes back to haunt me every once in a while --- where
software turns the shift key on and won't turn it off. Plus,
applications crash regularly, whether they be Internet Browsers (IE and
Netscape both), NFS (JoeK found this lovely way to crash his OS just
doing a file transfer), Xservers, Telnet terminals, or Pointcast...

Emacs has never crashed on my NT box, but I'm afraid to put any real
data on the machine for fear that it will get clobbered in a hard disk
crash, so I do all my work remotely -- my machine evolved over time and
paranoia into a glorified Xterminal.

> My point is that, in my experience, if an NT machine crashes, it's a
> hardware problem.

Yup, so I'm moving to an OS that understands the hardware, but given my
druthers, I'm much rather have an Alpha than an Intel.

Unfortunately, these machines were donated. If I had to buy, I would
buy an Alpha or a Mac.

Darn, check out the cool specs:

TWICE the performance of Pentium IIs at half the price. Runs
Intel-running software in emulation mode faster than Intel-running
software running on the Pentium IIs in native mode. Runs NT or Linux
natively. Designed for 64-bit, realtime MPEG1 authoring, host-based
DVD/MPEG2 playback *drool*

Forget more Suns, JoeK, we should have our next round of buys be Alphas.
Actually, I wouldn't mind a Mac either; DanZ has showed me some really
cool stuff, and I just know Rhapsody is going to be a wet dream.

Man, DEC's marketing department sucks. Alpha should just secede from
DEC, start its own damn swiftly moving Silicon Valley startup, and toast
Intel to the bone!!!! I would *love* for Andy Grove's last sight to be
that of his own heart sitting in the hands of an Alpha startup, held up
in front of him just as it's been ripped out of his chest, just so he
can see how truly black his heart really is...

Rivers of blood!!! Death to the infidels!!!! Yeah, yeah, kill, KILL!!!!

> So what the heck are YOU guys doing to make your NT machines crash???

Running xterms and xemacs and an occasional netscape or ghostview.
Every now and then running the Java runtime.

> I am really truly puzzled and I'm close to flying down there and setting
> up a debugger and trying to figure out what the problem is...



Perhaps the most controversial portion of Myhrvold's defense of the
[positive feedback cycle] theory was when he observed that, in the
history of computers, the market share leader in operating systems gets
about 90 percent of the market, the runner-up has about 90 percent of
the remainder, and so on. In applications software that runs on top of
the operating system, the positive feedback cycle is at work, too, but
it does not bring customers as many benefits, so the distribution is not
quite as dramatically skewed. Nevertheless, a pattern is found in
application software categories, too: the leader gets 60 to 70 percent
of the market and the runner gets 60 to 70 percent of the remainder.
Two years after Myhrvold made these observations, another software
category --- the suite of productivity applications --- had become
important. When a well-integrated suite like Microsoft's Office became
widely adopted, it brought out every one of [Stanford economist Brian]
Arthur's five sources of tipping, enhancing the effects of the positive
feedback cycle. By 1995, Microsoft's Office had a 90 percent share of
the suite segment of the market, exactly what Myhrvold's earlier remarks
had predicted.
-- Randall E. Stross, _The Microsoft Way_, p.187