The Lmerging Lconomic

Paradigm oI Open Source

The economics of Open Source can be explained entirely
within the context of conventional open-market econom-
ics. Open Source has much stronger ties to the phenome-
non of capitalism than many have appreciated.
Timeline
1981
Joins the New York Institute of
Technology, becomes an Unix ker-
nel programmer
1986
Works at Matrix Instruments for 6
months due to non-compete clause
1987 to 1999
Joins Pixar Animation Studios as a
senior systems programmer
1987
Initial release of Electric Fence
1994
Became a major Linux developer
1998
Starts the Open Source Initiative
with Eric Raymond
1999
Linux Capital Group
2000
Senior strategist for Linux and open
source at Hewlett-Packard
2002
Starts Perens LLC
2004
Senior Scientist, Ceorge \ashington
\niversity
Boards
SoItware in the Public Interest
Llara Fusion
Open Source Risk Management
Background
Bruce Perens helped coin the term Open Source" and wrote
the original Open Source dennition. He has been instrumental
in many open source initiatives, such as the Linux Standard
Base.
Bruce is series editor oI the Bruce Perens' Open Source Series
line oI books with Prentice Hall PTR publishers, with , titles
published so Iar. The books sell well in retail stores, even
though their texts are released under an Open Source license.
He was recently named Senior Scientist Ior Open Source by the
Cyber Security Policy Research Laboratory oI Ceorge \ashing-
ton \niversity.
Perens is a member oI the board oI directors oI Open Source
Risk Management Inc. and Llara Fusion and a major stock-
holder oI Progeny Linux. He is an elected director oI SoItware
in the Public Interest.
Introduction
Bruce Perens was SLForum's distinguished speaker at PARC on
April z, zoo,. Andrew Aitken, Iounder and managing partner,
Olliance Croup introduced Bruce as being a Iounder and driver
oI the Open Source movement. He observed that Bruce has a
tremendous sincerity and Iocus on Open Source and has consis-
tently Iocused on what is the next step to take in moving Open
Source Iorward.
Andrew started selling Open Source soItware, 8 or j years ago.
In attending many oI the same conIerences as Bruce, since they
both liked to drink beer, they oIten ended up at the same bars
aIter a conIerence. \hen Andrew was in a cab without cash,
Bruce handed him $zo without a question, hardly knowing him.
A Brief History and Background
In Andrew's introduction, he noted that Bruce
loaned him $zo, without knowing him, when
needing to pay a taxicab bill. But in reality,
Bruce had seen Andrew at conIerences beIore,
and knew he could nnd him again!
Bruce has done a bit oI everything. In being
asked to speak tonight, he reminded the audi-
ence that another term Ior distinguished
speaker is old Iart."
Bruce's programing career started when he was
running a college radio station. Having access
to a Xerox Sigma j that ran BASIC, he spent
six months nguring out how to use it to auto-
matically generate the radio station log. Being
particularly good at reading books, his next
task was to learn APL on the Sigma j.
Bruce's involvement with computer graphics
started in j8 when he joined the predecessor
to Pixar, the ew York Institute oI Technol-
ogy's computer graphics lab as a system opera-
tor on its \nix system at $z.o/hour in j8.
They were the nrst to use computer graphic
character animation Ior story-telling charac-
ters.
Having never taken any computer science
courses, at the time, there were only two books
on \nix. There was a pile oI supplemental
\nix literature that you had to print out on a
Versatec wet toner printer. And there was
Lennis and Ritchie's C programming book.
ow the reason that you had to know all this,
was that at the time, when your nle system got
corrupted, you had to use the ALB debugger
to nx it. \nix tools like Isck (disk check) didn't
exist at that time.
In j8, a single Irame player didn't exist. A
Irame buBer Ior a 6o x 8o image cost
$;o,ooo. Bruce worked Ior 6 or ; years on
LLC PLP- and VAX computers running
\nix. They developed soItware that let them
use a videotape recorder one Irame at a time,
down to a oth oI a second. This was his nrst
exposure to what you could do in realtime with
\nix.
He leIt the ew York Institute oI Technology
in j86 and went to Matrix Instruments (now
part oI AgIa) Ior six months, beIore going to
Pixar. At the time, there was a non-compete
clause that prevented him Irom going directly
to Lucasnlm that owned Pixar at the time.
In j86, Pixar's primary business was making
an image processing computer. Ceorge Lucas
needed money to make the Howard the Duck
movie, so he sold Pixar to Steve ]obs. At any
rate, the Pixar computer was a SIML (single
instruction, multiple data) computer with a
tiled memory architecture. It was imple-
mented using multiple -bit AML zjo proc-
essors. Bruce wrote microcode Ior the znd
generation oI this hardware.
They used gate arrays Ior part oI Pixar's hard-
ware. The problem was that it took or z
months to make a gate array and it cost
$,o,ooo each time you made one. They ad-
dressed this problem by creating a behavioral
model oI the gate array that allowed them to
simulate it, and get it right the nrst time.
Getting Involved with Open Source
Bruce got an email that took him about a halI
hour to answer a question about soItware that
he had written and made available on an Open
Source basis. He got an email back saying that
thanks to his advice, the code would be used
on the next space shuttle Bight.
Being a bit oI a loud mouth, Bruce enjoyed
talking to the press and became well known.
MicrosoIt started saying some really stupid
stuB to the eBect that Linux was a cancer, and
he responded to their accusations. He started
to get invitations to speak at conIerences, at
desirable locations like Iceland.
Bruce released his Llectric Fence program that
enabled the Malloc debugger to stop on the
exact line oI a bug, as Open Source in j8;.
The Iollowing things happened:

\hen it was nrst used on the HP-\X

system, it helped nnd o bugs, Iorcing an
unscheduled bug nx release.

Bruce got an email Irom a person in Ire-

land saying that the program had saved
their job.

It made money Ior him, when a person

emailed him all the documentation Ior
Llectric Fence, meaning that he didn't
have to spend time documenting it.
These and other events reinIorced Bruce's in-
terest and involvement in Open Source, even-
tually leading to his Iormation with Lric Ray-
mond oI the Open Source Foundation in jj8.
Subsequently at Pixar, he read a statement by
Steve Ballmer oI MicrosoIt stating that Open
Source soItware is a way oI licensing soItware."
At that point, Bruce realized that Steve had
read his open source dennition, and the inBu-
ence he was starting to have on the industry.
Open Source Economics
The tremendous economic success oI Micro-
soIt with its $oB in revenues and ,,,ooo em-
ployees, Irequently causes us to think oI soIt-
ware in a vendor centric mode. And it shapes
our thinking, causing us to wonder how people
make money doing Open Source: \e Iorget
that only o% oI soItware is sold as soItware.
Most soItware is written by customers.
Most companies make products and provide
services, only a small minority manuIacture
soItware. Programmers and web designers are
an unavoidable cost oI doing business Ior all
but soItware companies
The economic impact oI MicrosoIt or any
soItware company is in providing the tools that
enable modern businesses to operate.
Lespite widespread dissatisIaction with its
products, the lack oI competition due to its
being a monopoly, gives MicrosoIt no incentive
to improve its products. Open Source soItware
represents the nrst serious competition to Mi-
crosoIt in a decade in that it both democra-
tizes and provides economic incentive to keep
soItware competitive.
Open Source soItware has much stronger ties
to capitalism than many realize. The problem
with the book, The Cathedral and the Bazaar,
written in jj8, is that it describes program-
mer motivation Ior doing open source as being
non-economic in nature. \e now know that
this is not correct.
The Iundamental economic benent oI Open
Source is that it enables you to spend your
time and money doing something else - it
makes you more productive. As business peo-
ple get involved with Open Source soItware,
they are increasingly coming to appreciate this.
Differentiating Software
There are two types oI soItware:

LiBerentiate

on-diBerentiating
LiBerentiating soItware makes your business
better than the competition in some way:

Amazon's recommendation Ieature sug-

gests books that other people purchased
when you buy a book.

Pixar's internally developed soItware used

to make their nlms.
LiBerentiating soItware provides an advantage
over competitors and thus must remain unique
and proprietary.
Most soItware is non-diBerentiating. \hile it
is essential to your business, it does not make
your business any better than your competi-
tors. Two examples are:

Apache web server

COML desktop
Since non-diBerentiating soItware is common
to your competitors, making it open source
enables them to become your best collabora-
tors. This is why IBM and HP cooperate on
many Open Source programs.
Open Source Marketing
II you ask someone, would you liked to work
in a planned economy:", the answer invariably
is, no". \ith Open Source, you don't have the
situation oI large companies like MicrosoIt,
Oracle or SAP telling the entire IT industry
what to do.
You can think oI Open Source soItware as be-
ing a massively-parallel drunkard's walk nltered
by a Larwinistic process. \hen you have a
large collection oI people writing soItware,
some oI them are going to have good ideas.
The market decides what are good ideas and
what are not. This process works as well as
marketing.
\ith Open Source soItware, iI there is a group
as small as ,o people worldwide who want the
same thing, willing to work on it in their spare
time, you have suBcient resources to develop
large and complex soItware products. As the
product matures and becomes oI more interest
to people, the number oI people contributing
to it will grow substantially.
How Software is Developed
There are Iour ways in which soItware is de-
veloped:

Retail

In-house / contract

Consortium

Open Source
Retail soItware accounts Ior about o% oI all
existing soItware. You might think that it is a
very eBcient way oI distributing the cost oI
development among many customers.
But in Iact, even Ior MicrosoIt, in zoo it
spent only 6.8% oI its revenues on develop-
ment. AIter the middlemen takes their cut, and
nguring that ,o% oI all soItware ends up as
shelIware (it never gets used), only about ,
cents oI your soItware dollar actually went to
developing it.
\hy is this: Retail soItware has a very large
overhead, since it is very expensive to locate
customers, you have to manuIacture boxes and
pay Ior shelI space. There is advertising and
points to the distributor and retailer.
The other characteristic oI retail soItware is
that it requires a mass market so everyone can
buy it. The result is that retail soItware is non-
diBerentiating.
In-house and contract software avoid
much oI retail soItware's overhead, with the
result that you obtain about ,o% eBciency on
the dollar, and you gain a diBerentiating advan-
tage. Typically a single customer pays Ior eve-
rything, and takes all the risk. Historically
about ,o% oI these eBorts ends up as shelI-
ware, so the net is that you only get about z,
cents on the dollar Ior your programming.
Consortiums or Iormal corporate collabora-
tions have ended up as colossal Iailures, as wit-
nessed by the hundreds oI millions oI dollars
wasted by Taligent and Monterey. The Iatal
Baw oI consortium projects is that there's al-
ways the handshake with one hand and a dag-
ger in the other - it is never Iair Ior all oI the
partners.
Open Source has its initial development by a
single entity. It is released as soon as it does
something useIul. For example, Cnome ini-
tially used the user interIace Irom a paint pro-
gram, and could be booted and run Irom a PC.

So what does Open Source soItware do eco-

nomically:

At an very early stage, no one bares the

entire cost and risk oI development.

II I'm internally writing soItware that is

non-diBerentiating, than I'm wasting my
money since it isn't giving me a competi-
tive edge. II I'm going to write soItware
that is only Ior my own use, I want to Io-
cus on soItware that diBerentiates itselI.
This is where Sun went astray as a developer oI
non-diBerentiating system soItware. They were

There are some exceptions, it took Mozilla years beIore its release as FireIox.
Iat and happy making money with hardware
margins oI ;o%. But now that they have to
compete with Lell (who acts as a Iront Ior
Chinese manuIacturers) over halI a margin
point, they are in big trouble.
This is the same problem that HP has. In driv-
ing by HP's headquarters on Page Mill on the
way to this talk, I shouted out the window, I
told you turkeys that buying Compaq would
not be a good idea!"
Capitalism and Open Source
Steve Ballmer oI MicrosoIt has said that,
Open Source soItware is a threat to the econ-
omy." He must have cut his economics class at
StanIord.
At the beginning oI the zoth century, there was
a large industry devoted to the harvesting,
storage, and transportation oI ice. \ith the
development oI reIrigeration, this industry col-
lapsed. Lespite the Iailure oI this industry, it
was a tremendous boost Ior the overall econ-
omy, since it now worked more eBciently.
Although Open Source is not good news Ior
retail soItware companies, the overall demand
Ior soItware will not decrease. Instead retail
soItware programmers will be Iound working
internally at companies like Amazon and LBay,
developing soItware that provides them with
diBerentiating advantage.
By enabling things to be run more eBciently,
Open Source reduces the overall cost oI soIt-
ware, with the result that more people will be
able to aBord computers.
From a business standpoint, prontability can
be increased by spending less on cost centers
such as soItware. Open Source helps a business
maximize the money it spends on diBerentiat-
ing soItware.
\ith Open Source soItware, there is no man-
date over how people use a product, it provides
a level playing surIace Ior competition among
soItware vendors. This in contrast to Micro-
soIt who uses network eBects" to bias the
market.
What Will Stay Proprietary?
There is proprietary soItware that will proba-
bly not become Open Source. For example,
consider TurboTax. There is huge liability iI
you are wrong, and nobody is going to write it
Ior the love oI doing it. \hile it is quite com-
plex, it is also soItware that is only used once a
year.
(Bruce joked that although his wife uses Windows,
he does NOT undergo ritual purication befor
touching her PC to run TurboTax once a year whe
doing taxes!)
Another example is Synoptics, who provides
CAL soItware Ior designing integrated cir-
cuits. \hile there is a vigorous open source
hardware eBort, Synoptics sill has a several
year lead over these eBorts.
But Synoptics decided to make a key part oI its
soItware Open Source, soItware that enables it
to interIace to multiple chip Ioundries.
By giving this soItware to its competitors, it
created a common API and deIacto standard.
The result is that everyone uses it, and Synop-
tics no longer has to create unique interIaces
Ior each new Ioundry.
Economics of Finding Bugs
There are diBerent motivations and economics
Ior nnding and nxing bugs. For MicrosoIt, out-
side people are continually looking Ior bugs so
they can write viruses and worms. \ith Open
Source soItware, people are looking to extend
and modiIy it.
\hen Borland made Interbase Open Source,
nine months later, a major backdoor bug was
Iound. Here was soItware that had been used
to implement airline reservation systems, with
a bug that could be used to obtain Iree Bights.
This illustrates the advantage oI having many
eyes looking at code. However, many eyes are
not disciplined eyes that will inspect code in a
systematic manner - you need a mix oI the
two approaches.
Programmers have a lot in common with art-
ists. OIten the creative work that both artists
and programmers do is not specincally con-
nected with their employment. And like an art-
ist, Open Source programmers have a tremen-
dous pride in their reputation. This is why the
Linux kernel and other Open Source soItware
keeps getting better and better.
The Future of Open Source
Open Source soItware is enabled by the Inter-
net, that has essentially eliminated the cost oI
manuIacturing and distributing soItware.
\hen a Star Trek replicator becomes available,
design will become more important than
manuIacturing Ior other products as well.
The closest thing that we have today in Open
Source hardware is the Open Source soItware
radio receiver. It lets you receive a signal be-
tween 8oo MHz to CHz. \sing Python, you
can connect various LSP blocks to implement
a 8oz. radio, a HLTV decoder, or do radio
astronomy.
Open Source Opportunities
Linuxcare was a $zoM Iailed startup. Its Iun-
damental problem was that it was ahead oI its
time. Larly adopters oI Linux had no need Ior
support, they wanted complete control. But
the second generation oI Open Source users
want support and are willing to pay Ior it. This
is the opportunity that SpikeSource is seizing,
and it also represents an opportunity Ior both
IBM and HP.
The computer should be your invisible Iriend.
\e already have soItware that enables con-
tinuous speech recognition and generation.
\hat is needed are the higher levels oI soIt-
ware to make these capabilities useIul to us.
Fundamentally, it is no longer suBcient to just
know computers, you have to have domain ex-
pertise in something else.
Summary
\sing Open Source soItware enables a com-
pany to become more prontable, and provides
it with more control over its business. Open
Source allows a company to Iocus its nnite
soItware development resources on capabilities
that diBerentiate itselI Irom competitors.
For More Information
Bruce Perens's paper, The Emerging Economic
Paradigm of Open Source can be Iound at:
www.perens.com/articles/economic.html
About the Author
Mark Luncan is a marketing consultant who
Iocuses on emerging technologies, assisting
companies in entering new markets and devel-
oping new business opportunities. He can be
contacted at markCaskmar.com.