[FSF India] [ARTICLE] Microsoft and GPL

Mon, 6 Aug 2001 10:12:59 +0530 (IST)

Wednesday, June 27, 2001 | Last Updated 11:19am

Why is Microsoft Attacking the GPL?

by Bryan Pfaffenberger <bp@virginia.edu>, 27-June-2001

Microsoft sees a threat to its underlying business model, which explains
the virulence of this particular round of FUD.

First it was "un-American"; now it's a "virus", a "cancer" and -- are you
ready for this -- "Pac-Man". Microsoft's high-profile, expensive and
carefully planned attack on open-source software is swinging into high
gear; so is speculation about exactly what the company hopes to achieve.
Typical is the conclusion of an analyst quoted by CNET News.com: Linux has
emerged as a "spoiler that will prevent Microsoft from achieving a dominant
position" in the server operating system market. By means of an especially
virulent FUD (Fear, Uncertainty, Doubt) campaign, Microsoft hopes to scare
companies away from Linux and toward Microsoft's newest offering, Windows
XP.

So far as it goes, this analysis is surely correct. Viewed from the
perspective of business history, however, this FUD campaign seems
mysteriously out of proportion with the threat actually posed by its
supposed object, Linux. Sure, Linux is a threat, and that's especially true
now that Linux is being so strongly backed by Microsoft's remaining market
opponents, including IBM. But the FUD campaign isn't directed against
Linux. Aside from occasional digs at Linux, the campaign is directed
against the licensing scheme, called the General Public License (GPL), that
Linux uses. To be sure, the analysts have an explanation for this, too. The
GPL is controversial even within the Open Source community, so you attack
Linux by attacking the GPL. But I think there's more to it.

Here's my argument, in a nutshell:

Business history teaches the following lesson: When a market-dominating
firm engages in a FUD campaign of this magnitude, it's not merely because
they're scared of competition from a new market entrant. Often, it's
because the new market entrant is seen to challenge the business model that
has enabled the market-dominant firm to make huge gobs of money. I believe
the GPL does pose a threat to Microsoft's business model, and that's why
the free software licensing scheme is under such concerted attack.
Specifically, the GPL threatens Microsoft's ability to preserve what
economists and legal scholars (as well as the judge in the Microsoft
antitrust case) call the "application barrier to entry" -- the primary
means by which Microsoft has been able to establish and preserve commanding
dominance in its core markets.

I'll make this argument by recounting the story of the archetypical FUD
campaign, IBM's 1970s-era effort to discredit the upstart Amdahl
Corporation. I'll also examine some of the evidence that's come to light --
some of it from the lips of none other than Gates himself -- since
Microsoft's FUD campaign started. Let me apologize in advance for the
length of this essay; the issues are too important to be glossed over. But
I've included lots of subheadings so you can skim around, if you like.

Classical FUD: IBM vs. Amdahl

When ex-IBM executive Gene Amdahl founded Amdahl Corporation in 1970, he
reflected gloomily that the new company would soon become the target of an
aggressive IBM marketing strategy, which Amdahl termed "fear, uncertainty,
and doubt" (FUD). As subsequent events affirmed, Amdahl's prediction was
right on the money, and IBM's FUD campaign nearly put Amdahl out of
business.

IBM in the 1970s was huge profits with little incentive to innovate. Here's
the background. IBM's mid-century executives, like those of other
manufacturing firms, knew perfectly well that maximum profits follow from a
"leasing-only, no-sales" policy; as economists have since theoretically
demonstrated, such a policy maximizes revenues because it keeps secondary
goods off the market. But US antitrust prosecution, culminating in a 1956
consent decree, forced IBM to sell as well as lease its computers.

For federal regulators, the consent decree was a hollow victory. IBM neatly
side-stepped the regulatory framework by means of measures such as net
pricing, in which the cost of IBM installation labor was bundled into the
price of IBM computer systems; customers who wished to purchase non-IBM,
third-party peripherals were thus forced to pay two installation fees --
which meant that most of them chose to stay with IBM's equipment. And when
IBM installed new equipment, the company generally required its customers
to surrender the old equipment for "rehabilitation".

Measures such as net pricing were combined with a quiet, low-level FUD
campaign, which was carried out by the firm's salespeople while they were
in the field. IT managers were told, "Sure, you can switch to the
competitor's product, but we can't guarantee your system any more, you'll
have to pay the highest per-incident charges for repairs, and we can't
guarantee that our components will function correctly in the future." In
other words, you're taking a dive off a cliff into the unknown, and it's
going to cost your company money, and later it may cost you your job. These
well-honed tactics led more than a few computing managers to agree with the
famous dictum, "You'll never lose your job by choosing IBM."

By means of net pricing and FUD, IBM managed to side-step antitrust
regulation and retained control of the secondary market. With almost
obscene profits and rapid growth, all was well at IBM, except there was
little incentive to innovate.

Amdahl Forms his own Company

As Amdahl watched IBM's processors fall behind the technical curve, he grew
increasingly concerned. At the same time that IBM was encouraging companies
to move more of their business operations to IBM mainframes, the company
was failing to deliver the processing horsepower that would be needed to
meet customer's needs. Amdahl proposed a series of new, higher performance
processors, but failed to win approval from IBM executives to develop them.
Intimately familiar with IBM's processor design, Amdahl knew that it was
possible to create a new generation of processors that beat IBM's products
in price as well as performance. So the young engineer quit IBM and started
his own company -- which still exists, incidentally, even though Amdahl is
no longer associated with it.

Amdahl's plans must have terrified IBM executives. Unlike IBM's ineffective
and increasingly marginalized competitors, Amdahl was intimately familiar
with IBM technology and proposed to manufacture IBM-compatible CPUs. If
Amdahl succeeded, IBM would be transformed, in effect, into a peripheral
manufacturer -- and with the loss of its control over the central
processing unit would come a concomitant destruction of its highly
profitable business model.

The Microcode Factor

What was wrong with IBM's processors? In case it has been a while since
your last electrical engineering class, here's the background. A computer's
central processing unit (CPU) is designed to perform a fixed number of
computational operations; the list of such operations is called the CPU's
instruction set. At the time, the fastest processors implemented their
instruction sets by means of "hard-wired" circuits, that is, circuits built
into the physical structure of the processor. When IBM designed the new
series of System/360 computers, the first product line to include a range
of compatible machines, the decision was made to implement the instruction
set by means of microcode.

In brief, microcode uses software, typically encoded in a bank of read-only
memory, to implement the lion's share of the CPU's instruction set. The use
of microcode enabled IBM engineers to design a single processor that could
be used in all the System/360 products, from small to large. The marketing
potential was enormous. For example, one of the earliest System/360
computers was offered with a microcode adjustment that enabled the computer
to emulate an earlier IBM machine, the 1401. However, the use of microcode
came at a price: diminished processor efficiency. Amdahl knew that he could
easily beat IBM's processor performance by creating processors hard-wired
with the System/360's instruction set. Amdahl Corporation's processors,
accordingly, did not use IBM's microcode architecture; instead, the firm
created less expensive and more powerful processors that gained economy and
efficiency from hard-wiring the instruction set into the CPU.

Employing an efficient, low-cost implementation of IBM's System/360
instruction set, Amdahl's first processors appeared in 1975; they undercut
IBM's price-performance points yet they were fully compatible with IBM
software, in which the firm's customers had made significant investments.
The result was tremendous enthusiasm among Amdahl's potential customers who
were quite ready to break with IBM -- even to the extent of paying IBM for
components and services that weren't actually received -- if the result
meant they had the processing power they needed. In 1976, Amdahl released
the V/6 processor, which offered three times the performance of IBM's most
powerful CPU.

FUD Kicks into High Gear

Faced with a threat to its underlying business model, IBM took its FUD
campaign to an unprecedented level, one that involved public statements, a
host of new product announcements and technological changes that seem to
have been aimed squarely at Amdahl's upstart processors.

In 1977, just as Amdahl Corporation seemed poised to take over a
significant percentage of the processor market, IBM announced the 3033, a
processor that -- IBM claimed -- would go head-to-head with the best Amdahl
could offer. But IBM did not merely announce a new processor. The firm also
announced revisions to the System/360 instruction set. In attempting to
address the Amdahl factor, IBM scientists discovered they could improve
processor efficiency by adding fourteen instructions to the System/360
instruction set. (This is quite an interesting "discovery," since it has
been proven time and again since then that reducing the instruction set,
not expanding it, is the best way to improve processor performance.)

IBM executives and engineers vehemently denied that the microcode changes
originated with Amdahl in mind, but the changes certainly had an
anti-Amdahl effect. The planned changes meant that Amdahl's hard-wired
processors would not continue to be compatible until Amdahl discovered the
changes that IBM actually made and figured out a way to implement them. At
the worst, Amdahl would have to go through the very expensive and
time-consuming process of redesigning his processors. For its part, IBM
made it abundantly clear to all concerned that the firm had no intention of
disclosing the technical specifications of the fourteen new instructions.
Amdahl would have to wait until the 3033 appeared, when he could legally
obtain one of the processors and discover the nature of the new
instructions through reverse engineering. From that point, it could be
months or years before Amdahl would be able to release a new processor that
would be compatible with the 3033. As if to pour salt in the wounds, IBM
warned darkly, and publicly, that the firm felt quite free to implement
additional microcode changes whenever doing so made sense for its
customers.

The 3033's preannouncement took the wind out of Amdahl's sails. Orders
plummeted, and the few remaining customers demanded contractual guarantees
that Amdahl's processors would remain compatible with IBM's. Most customers
decided to sit tight until the 3033 hit the market, one year later.

When the 3033 was released in 1978, the processor turned out to be inferior
to Amdahl's V/6, but it did offer better performance than previous IBM
processors, and it was an IBM processor. The pent-up demand for
high-performance processors unleashed a deluge of orders, to the tune of $6
billion, and IBM was forced for the first time to develop a lottery program
so that it could parcel out the scarce processors fairly. Although the new
processor turned out to be technically inferior to Amdahl's products -- in
fact, IBM's processors did not equal Amdahl's price/performance points
until five years later -- Amdahl's sales declined sharply. By 1979, Amdahl
was but a niche player in the mainframe industry, surviving only by means
of a capital infusion from Fujitsu, Inc. Meanwhile IBM's System/370
mainframes had captured 97 percent of the market.

Why the GPL Terrifies Microsoft

My analysis of the 1970s IBM campaign against Amdahl suggests the
following: FUD campaigns step up in virulence when the FUD-perpetrating
firm sees a threat to its core business model.

What's Microsoft's core business model, and how could the GPL, a mere
licensing scheme, pose a meaningful threat to such a gigantic, powerful
corporation?

My take on Microsoft's business model is, in brief:

* * Microsoft's core business model relies on an at-all-costs defense of
its overwhelming market dominance in end-user operating systems, because
this dominance is the lever by which the company hopes to achieve all of
its myriad other ambitions, including its ambitions in the server market.

* * To defend its overwhelming market dominance in end-user operating
systems, Microsoft must discourage or prevent the formation of a
critical-mass pool of non-Microsoft end-user applications. People don't
switch to a different end-user operating system until there's a
sufficiently large pool of applications from which to draw.

* * To prevent pools of non-Microsoft applications from forming, Microsoft
likes to appropriate what it calls "commodity protocols" (off-the-shelf,
public protocols such as HTML, JavaScript, CSS and many more), and add
proprietary extensions that prevent the formation of competing application
pools.

To grasp this model's implications firmly, consider Judge Jackson's
findings in the Microsoft antitrust case. Jackson did not rule against
Microsoft -- as so many people unfortunately think -- because Microsoft
made a better browser, gave it away for free and trounced Netscape in the
marketplace. Sure, Microsoft wants the browser platform, for all kinds of
reasons. But what really terrified Microsoft executives, I believe, were
the all-but-forgotten, pie-in-the-sky plans that Andreesen & Co. once made
to use Netscape as the foundation for a new, network-savvy, non-Microsoft
operating system and application pool. (I'll bet you didn't know this, but
Netscape came up with what amounts to the .NET idea long before it ever
occurred to those master innovators in Redmond.) Equally terrifying to
Microsoft executives was Java, and for the same reason -- it could be used
to create a non-Microsoft application pool. In both cases, noted Judge
Jackson, Microsoft used "embrace and extend" not only to discourage a
competitor but, more importantly, to prevent the formation of a competing
application pool.

Why the GPL Threatens Microsoft's Core Business Model

Still with me? Here's why Microsoft is attacking the GPL:

* * Microsoft can't play its "embrace and extend" game with GPL-licensed
software because the company can't appropriate and modify the code. If
Linux had been released under the BSD license, Microsoft would have
probably already released a version of Linux, Linux++ or Linux# or L-Nux,
with a variety of maddeningly incompatible oddities that taken together
would make it even more difficult to develop applications for Linux.

* * A GPL-licensed application pool is indeed forming around Linux, and
Microsoft can't figure out how to attack it. You can't attack the
companies, because -- as Eazel recently proved -- the software's still
around, even if the company shuts down or gives up on the product.

From the Microsoft perspective, GPL-licensed software is like those
monsters from "The Night of the Living Dead": they just keep coming back at
you.

Doubtful? Read this:

Mr. Gates made the following statement last week to a CNET News.com
reporter: "The ecosystem where you have free software and commercial
software -- and customers always get to decide which they use -- that's a
very important and healthy ecosystem", Gates told the interviewer. But the
GPL, Gates says, "breaks that cycle -- that is, it makes it impossible for
a commercial company to use any of that work or build on any of that work.
So what you saw with TCP/IP or Sendmail or the browser could never happen.
We believe there should be free software and commercial software; there
should be a rich ecosystem that works around that."

If Mr. Gates can forgive me for putting words in his mouth, here's my
translation: "There should be free software that we can appropriate and
modify -- we just love BSD stuff -- as well as Microsoft software. That's
very healthy (for us), because we can use this system and our embrace and
extend tricks to keep competing application pools from forming. But we
can't do that with GPL-licensed software. GPL-licensed software is not
enriching for Microsoft; it scares the living daylights out of us, in
fact."

In the coming weeks, you're likely to see Microsoft pressure to force the
U.S. government to disallow the use of the GPL as a license for software
created with public funds. If my analysis is correct, the decision should
go the other way -- the government should require anyone developing
software with government funds to release the software under the GPL. It's
the only way to ensure there's a meaningful public commons of freely
available software that can't be manipulated for predatory purposes.

Bryan Pfaffenberger is Associate Professor of Technology, Culture and
Communication at the University of Virginia, in Charlottesville, VA.