Re: Seeing VERSIONINFO under Vista?

ANd that's my point: their marketing department could predict in ten years how many
keypunches they would sell, but nobody foresaw the demise of the card as an input medium!
So when the assumptions on which the projections are based change, the projections become
meaningless.

NOBODY had a clue that the PC would become what it has. I was in the industry at that
time, and we knew it could never be more than a niche product. I was part of the 3M group
at CMU in 1979. "3M" was our designation for the ultimate personal workstation: a 1MIPS
machine, with a 1megapixel B&W display and 1MBye of memory. This was seen as the sort of
thing that a place like CMU would use for many years, because it pushed the envelope so
hard. My first PDA exceeded this in all dimensions except the display. My laptop exceeds
it by a factor of more than 1000 in memory and processor power, although onlly by a factor
of 20 for the display (1 megapixels of 24-bit color). A belief that IBM "should ought to
have had" people capable of seeing the future is asinine. They could at best do
extrapolation from the existing assumptions, but punctuated equilibrium invalidates
gradualism any time.

More below...
*****
On Wed, 30 May 2007 12:16:49 +0100, Daniel James <wastebasket@xxxxxxxxxxxxxxxx> wrote:

In article news:<63hp53pjml8l7qj8utalm3gsf83037t7mg@xxxxxxx>, Joseph M. Newcomer
wrote:
Not at all. This very interesting side-discussion started when you said
"Don't blame IBM". I'm just explaining why I do blame IBM.
****
For failing to have a crystal ball? That's what I'm objecting to!

No, not for failing to have a crystal ball -- IBM is (and was) a big business.
It has marketing and research departments that are surely equal to any but the
best of crystal balls. The problem was that IBM -- partly through arrogance and
complacence, and partly because of a lack of coherent strategy throughout the
organization -- didn't make use of the information that was available.

It's not that they didn't *have* a crystal ball, just that they didn't look into
it.

"Blame" implies a deliberate decision or set of decisions to build something
that had self-limiting problems.

I disagree. Blame can be ascribed for negligence, not just for deliberate acts.
*****
So I could be sued if a child climbs onto the roof of my house and falls off, because I
did not post a sign saying "Danger! Climbing to great heights is dangerous!" on all four
corners of the house?

Negligence assumes that there is a failure to take an action when the need for the action
is obvious. There was no need to do a massive market study for a market as small as the
personal computer market.
*****

However, I'm talking about "blame" because that's the word you used. When I say
that I "blame" IBM I don't mean to suggest that I, personally, have suffered in
any material way from their actions. I suppose I'm just saying that a number of
decisions made by various departments at IBM at and just after the time when the
first IBM PC was being designed and delivered have shaped the landscape of
computing for more than two decades ... and the responsibility for that must lie
with IBM.

There was no "different department" issue; there was ONE SMALL GROUP IN BOCA
RATON with a well-defined goal to produce an affordable personal computer ...

The "different departments" to which I alluded were the strategic planners and
marketing people who failed to pay proper attention to the PC because their
heads were up their corporate mainframes rather than looking at how the market
was actually moving. They didn't need a crystal ball to see that PCs were going
to be important, they just needed to open their eyes.
****
We could see where the market was going. It was going from mainframes to workstations.
Sun workstations, for example. A PC was not in the same class as a Sun workstation by a
couple orders of magnitude. So the people at IBM were predicting the future of
workstations, where it was obviously going to be an explosive market. Given the power of
a PC, they might as well have been projecting the market for abaci. It was seen as about
the same value. (This was wrong, but the whole point of punctuated equilibrium is that it
relates to catastrophe theory, which had little applicability in market research).
****

Whether they would have had the courage to bring a PC to market at all, if they
had looked at the market and understood what they saw I don't know -- but it's
pointless to speculate on that. If IBM hadn't made the PC someone else would
have, and history would be different.

... just assigning the blame where I see it as being due.
*****
Due to whom? This is where I think you are completely wrong. The concept
that there is even someone to "blame" is already a flawed concept, so
anything based on that fundamentally flawed assumption is nonsensical.

Lighten up, Joe! You're reading entirely too much into "blame".

This whole thing started because you wrote "Don't blame IBM" -- your remark
established the ground rule that the concept of "blame" is meaningful for the
purposes of this discussion. Don't worry, no-one here is going to nuke
Poughkeepsie as a result of it!

Even so, when the PC was new many people asked why the 8088 was
run at "only" 4.77MHz (IIRC intel sold it as a 6MHz part).
****
Oh, the reason for that is simple: the only crystal oscillator that was
available at the price and quantity required was one used for TVs, which
used a 4.77MHz oscillator!

Understood. I wasn't really looking for a reason, just pointing out that 4.77MHz
wasn't really such a big deal.

Clueless, certainly. The only ineptness was in not having an
open-collector/open drain interrupt line so that interrupts could not be
shared. But in that sense, nobody else of that era had a clue either, so
they were not alone.

I meant inept at marketing and corporate strategy -- or "crystal ball gazing" if
you prefer. <smile>

Absolutely. Motorola clearly screwed up; had they bid the 68000 to spec,
they would be the semiconductor powerhouse today, and Intel might not even
exist.

OK, I blame IBM *and* Motorola. Happier?

Note that comparing the 8088 to a Core2 or an XScale ARM is sort of silly,
since those are MUCH later architectures ...

You started it. I was responding to your:
What, by the way, is "crappy" about the chip? Right now, Intel produces
the fastest consumer chip on the planet.

I agree that intel NOW produces some very good chips, but we had been talking
about the 8088 ... which I still maintain should be judged as "crappy" by any
techical yardstick of its day.

The PC XT/370 was ...
[snip]
*****
Yes, I knew about that.

I thought you probably would have done.

One of the problems in comparing PCs to mainframes is that mainframes tend
to run massively-I/O-bound code, and until recently NO x86-class
architecture could come within a couple orders of magnitude of mainframe
I/O bandwidth. The NUMA x64 machines, however, have essentially destroyed
the advantage mainframes held there, and a machine equipped with a couple
$3000 Adaptec cards is now in the same performance range as many
mainframes,but is a fraction of their cost. So having a 68K-speed CPU that
ran mainframe programs made no sense, because a single-bus PC (even a
Microchannel architecture) simply did not have the I/O bandwidth to run most
mainframe apps.

A number of interesting points, there.

Yes, mainframes have much better i/o handling than PCs ... but N PCs handling
one task each can outperform 1 mainframe running N tasks with N sets of i/o (for
suitably large values of N). Where the break-even point comes depends a lot on
the size of the mainframe and the nature of the tasks being run -- my old
University bought an ICL 2900 mainframe (the government insisted that they buy
British, so there wasn't much choice) on the understanding that it could support
over 100 ussers working interactively at one time. It turned out that ICL's
understanding of "working interactively" meant "using terminals attached to one
monolithic application" -- as in data entry tasks) while the University's
requirement was 100 users each running sizeable programs in separate VMs.
Understandably, the system crawled. Eventually the government and ICL coughed up
a million or so pounds for the University to buy something else to handle the
heavy interactive use, and the University got a couple of VAX-11/780s. While the
VAXes couldn't crunch numbers as fast as the 2900 they gave a much more
responsive service when 100 people were online at once, and half of them were
running compilers.

As for a desktop machine running mainframe software making no sense ... it
depends what the intended use is. Most mainframe software only has high i/o
bandwidth requirements because it is doing i/o for a lot of users at once. A
desktop running mainframe software is likely to be being used by just one user
and won't have the bandwidth requirements. Something like the XT/370 strikes me
as ideal for a small consultancy offering software development and maintenance
services for mainframe code, but not having the need to process the volumes of
data that a live installation (on a real mainframe) would have to handle. It may
also be useful for small companies that want to process the same sort of data in
the same sort of way and with the same software as much larger companies, but
who don't have the volume of data (or the capital) to buy their own mainframes
(in this case it might be cheaper to buy time on a real mainframe, but the cost
dynamics probably vary form case to case).
*****
Part of the issue about bandwidth is that the reason we put 100 people on the system is
because it puts them close to the data. So while no one user impacts the bandwidth, 100
users do. The alternative is to replicate the data, and these days we are looking at
terabyte databases being the state-of-the-art. Replication costs are high because of the
cost of getting the data to the site (yes, a 1TB drive is $400, but how do you load it? Or
keep it in sync with all the other machines?)

But as I indicated, that is becoming less relevant, as we can afford to do high-bandwidth
servers that are closer to the end users.
*****

And a FORTRAN programmer really doesn't care what architecture is executing
FORTRAN.

As long as all interaction with the real world is abstracted by the runtime
libraries that's probably true, but as soon as he has to use some
platform-specific API he's going to need to test it on the right platform.
*****
That's why I chose FORTRAN (or COBOL) because their interactions with the system are
governed by standards. Similarly, C using only the standard C library or C++ using the
standard C++ library, can be made 100% portable to any environment.
*****

Supercomputers do have their uses, but there aren't very many of them.
They're still expensive, and you can really only justify them if you have
massive budgets and massive computing need. Otherwise, arrays of PCs will
do the job.

Well, it's certainly true that you can't justify them if you can't afford them.
The interesting question comes when you /can/ afford it, and is about the
position of the break-even point between supercomputers and clusters of big
workstations -- and that all depends on the sort of work you need them to do.

IBM still offer mainframes, and one of the purposes that they offer them for is
running large numbers of VMs hosting alien OSes (such as linux). In some cases
the mainframe can work out more economical than discrete workstations
(especially when the workstations are typically not all used at once, so
load-balancing is possible) and in others the workstations work out cheaper.

Then again: A friend of mine works for IBM, and a lot of his work is at the
ECMWF where they have clusters of supercomputers ... I don't see that lot being
replaced by a couple of PCs and bit of wet string any time soon!

http://www.ecmwf.int/services/computing/overview/index.html
Joseph M. Newcomer [MVP]
email: newcomer@xxxxxxxxxxxx
Web: http://www.flounder.com
MVP Tips: http://www.flounder.com/mvp_tips.htm
.

Relevant Pages

  • Re: Seeing VERSIONINFO under Vista?
    ... no desire to impact upon the marketplace for their mainframe computers. ... NOT be a serious threat to the main business. ... IBM, which is what we were talking about. ... It was a SMALL market. ...
    (microsoft.public.vc.mfc)
  • Re: The most dangerous product the IBM mainframe has ever seen
    ... >recognition of the IBM mainframe as a major player even today. ... Don't confuse high profit with high market share. ... IBM still derives 70% of its Net Profit from mainframes. ... Even contracting managers. ...
    (comp.lang.cobol)
  • Re: Another one bites the dust
    ... A real killer was that when IBM started to charge for software, ... in question to stay on the mainframe; ... educational mainframe market a long time ago. ... but I think there was some kind of consent decree or other ...
    (bit.listserv.ibm-main)
  • Re: Press Release: IBM Academic Initiative Boosts Mainframe Community, Helps Plug Projected Skills G
    ... > IBM Academic Initiative Boosts Mainframe Community, ... > eServer zSeries mainframe through its Academic Initiative. ... and training for students and faculty. ...
    (bit.listserv.ibm-main)
  • Re: Is a Hurricane about to hit IBM ?
    ... We tried WebSphere on the mainframe and it is too slow ... Compare the investments IBM is ... -EVERYONE- who runs a TPC-C benchmark fudges the hell out of it. ... These days the machines are so powerful most customers are pretty happy ...
    (bit.listserv.ibm-main)