2003-12-05 16:35:35 UTC
by David S. Evans, senior vice-president at National Economic Research
How's this for a way to organize a market? Let's all agree to work together
to create a new product. I'll put in some effort, and hope you and many
others do, too. Once we've got a working prototype, we'll let anyone use it
for free. Of course, no one can possibly make money from the enterprise. But
if everyone works together, we'll have some fun and end up with a great
product we can all use.
Your typical economist would probably assume that she was listening to some
latter-day Rip van Winkle who didn't know that the events of the last decade
or so-notably, the collapse of communism-had given utopian dreams a bad
name. A more patient one might resist the inclination to make a dash for the
exit and explain to our Haight-Ashbury refugee about the importance of
property rights, free-rider problems, and profit incentives in making things
people really want at a cost that reflects the value of the labor and
capital that go into them.
That economist, however, won't have any easy job responding to the retort:
so where do the Linux PC operating system and a host of other distinguished
software coming out of the open-source movement fit in this picture?
Linux, you might not recall, was not so long ago just a gleam in the eye of
Linus Torvalds, a computer science student in Finland. He posted an early
version of his nascent software code for a new operating system on the
Internet and invited others to expand and improve it. Many programmers
responded with enthusiasm. During the 1990s, Linux became one of the leading
operating systems used for servers-the computers at the heart of networks as
small as the ones used by neighborhood hardware stores and as large as the
Linux, moreover, was not a flash in the pan. Several other software products
that are widely used on Web servers-Apache and Sendmail in particular-arose
from this anomalous cooperative approach. As with all "open-source"
software, anyone can get a copy of the underlying software code and, in most
cases, use it for free.
Open-source software thus presents a fascinating puzzle to economists. If
one described this method for producing software when Linus Torvalds was
just another aspiring computer nerd, most economists would have predicted
that it had zero chance of working-assuming they took the proposition
seriously enough to prognosticate at all. Now that open source has generated
unquestionable successes, economists are trying to understand why it works,
how it works, how it is likely to evolve, and what lessons it provides for
other products. Herewith, a primer.
Open-Source and Proprietary Software
Strictly speaking, "open source" refers only to the practice of making the
source code for software available to others at no extra charge, although
possibly with some restrictions on how they can use it or pass it on to
others. Source code is what programmers actually write, using various
programming languages. Linus Torvalds allowed anyone to use his original
Linux code, so long as they agreed to give others free access to any
subsequent changes they made to the source code. Or, to put it crudely: they
got for free and had to give back for free.
"Proprietary" software, by contrast, is sold like cars or telephone service.
The entity that owns the rights sells licenses to use their intellectual
property, guarding against theft either by not disclosing the source code or
by limiting the use of the source code enough to prevent the recipient from
Consider the most popular word processing programs over the 20-year history
of the personal computer, WordStar, WordPerfect, and Word. The vendors of
these programs distributed them to customers in the form of machine code
(not source code), under terms that prevented users from legally making
copies for others. Machine code (sometimes called binary code or object
code) is in a form that even experts cannot easily decipher. Vendors of
proprietary software, including such firms as IBM, Sun, Oracle, Apple and
Microsoft, also protect their rights by enforcing copyright and patent
The issue of property rights in software has been simmering since the dawn
of the computer age in the 1950s. Computer users wrote programs in the early
languages like FORTRAN that often came with the big computers, and often
shared the results with others. Likewise, many of the early personal
computer (PC) programs were created by hobbyists and distributed for free.
But, although "shareware" or "freeware" programs have remained popular, the
production of proprietary mass-market software exploded with the spread of
PCs through homes and offices. Indeed, proprietary programs were important
in most categories even when computers still numbered in the hundreds of
thousands instead of the hundreds of millions. Among the early best-sellers:
CP/M in operating systems, MS-BASIC in computer languages, WordStar in word
processing, and VisiCalc in spreadsheets.
The recent success of open-source software-and the buzz around this method
of production-resulted from several developments. First, several formal
institutions were established to promote open-source software, most notably
the Free Software Foundation in 1985 and the Open Source Initiative in 1998.
Second, open-source software became significantly easier to produce as a
result of the growth of the Internet and the reduction in communication
costs. Access to the pre-Web Internet in the 1980s was mostly limited to
university research centers.
Third, servers-the computers that do the background work for computer
networks-and the early Internet itself served as a primordial stew for open
source. Many of the successful open-source products to date have resulted
from the needs of "techies" who maintain servers. They want software with
the source code available so they can fix glitches themselves. These
successes include Linux (which accounted for an estimated 12 percent of
operating systems on new servers shipped in 2000), Apache (which by some
measures is now the most popular Web server software), and Sendmail (which
is used on a majority of Internet mail servers).
Software Production and the Role of Intellectual Property
In market-oriented economies, the owners of computer software can exercise
intellectual property rights to protect their work. The creator of software
can obtain a copyright, which enables the owner to limit the legal ability
of others to copy or modify the work. Moreover, in some countries the
creator can also obtain patents over techniques used in the software,
further limiting the legal ability of others to use the knowledge embodied
No less important, the software owner can maintain trade secrets much more
easily than for written works in general. Source code is written in a
computer language that is understandable to people trained in that language
[e.g. "if (myNumber < 25) myNumber = myNumber + 1")]. But before it can be
used, this source code must be translated into machine code that is
understandable to the computer hardware, but terribly hard for humans to
decipher. A program distributed in source code therefore can be readily
modified or adapted for other purposes by a skilled programmer, while a
program distributed only in machine code cannot.
Note that both proprietary and open-source software programs have copyright
(and sometimes patent) owners. In both cases, the owners license the use of
the intellectual property in their software to others. But the licenses
governing proprietary and open-source software use differ radically.
Proprietary software vendors use their property rights to prevent anyone
from using the software without paying for it. And they are not bashful
about going to court to challenge infringement. In perhaps the most famous
software copyright infringement case, Apple sued Microsoft to prevent the
use of a point-and-click graphical user interface in Windows (Apple lost).
Open-source adherents, by contrast, typically grant their intellectual
property rights for free to others. In general, an open-source license
permits anyone to take the software, modify the source code, and use or
distribute the modified source code to others without paying the owners of
the original copyright. One type of license goes further, requiring that if
a program is modified and distributed, the modified source code must be made
available for free. That license, designed by the Free Software Foundation,
enforces a kind of social compact among members of the open-source community
and helps to advance the FSF's goal of making software "free." Linux is
licensed this way, as, apparently, is most open-source software.
In a nutshell, then, proprietary software owners use intellectual property
rights to make profits by excluding non-payers. Most open-source software
owners use intellectual property rights to enforce agreements among
The proprietary model is easily understood-it's about making money. Firms
invest in the creation of software. They hope to realize a return on that
investment by licensing the software to consumers, and possibly by licensing
or selling the software code to other software firms. The open-source method
is not well understood, but is clearly rooted in a complex constellation of
incentives ranging from purely social (like participating in any club or
group activity) to the profit-oriented by firms (such as IBM) that see the
success of open source as advancing their own interests.
The proprietary model usually leads firms to organize software production
through familiar hierarchical methods. Typically, a firm anticipates
consumer demand for a particular software product (possibly educated by
consumer research). Software architects design the software, programmers
write it, testers identify bugs, and programmers fix those bugs. The major
creative work for any software module tends to be done by individuals who
are in close proximity.
Today's open-source model more typically relies on a decentralized
production-what Eric Raymond, one founder of the Open Source Initiative,
calls the "bazaar" as opposed to the cathedral. A common method is for a
project to have a central committee (as for Apache) or a benevolent dictator
(as for Linux) that determines the general software design, identifies
projects to be worked on, and determines what modifications get incorporated
in the "official" version of the open-source product. Individual programmers
pursue these projects and identify and possibly fix bugs in the software.
They may also work on projects of their own choosing that result in
modifications to the software. Communication is electronic, via email and
The FSF's General Public License
The Free Software Foundation opposes proprietary software on ethical
grounds. As Richard Stallman put it, "Even if GNU [his operating system
project] had no technical advantage over Unix, it would have a social
advantage, allowing users to cooperate, and an ethical advantage, respecting
the user's freedom." The FSF promotes what it calls "free software" and
disavows any support for the broader concept of "open source" software."
The FSF is therefore not seeking to coexist with proprietary software, but
to displace it. Although many other participants in the open-source
community do not share the FSF's strong views, the General Public License
(GPL) popularized by the FSF has been critical in shaping the open-source
software industry. The GPL relies on the legal tools at its disposal-the
rights of a copyright holder to decide how the copyrighted work is
licensed-to encourage the expansion of open-source software and discourage
the use of proprietary software with open source.
The GPL requires that anyone who wants to distribute modified versions of
software covered by the GPL must make the source code for the modified
software available at effectively no charge. According to the FSF,
We encourage two-way cooperation by rejecting parasites: whoever wishes to
copy parts of our software into his program must let us use parts of that
program in our programs. Nobody is forced to join our club, but those who
wish to participate must offer us the same cooperation they receive from us.
This forced sharing of modified code results from a feature of GPL dubbed
"copyleft." The gist is captured in the following license clause:
You must cause any work that you distribute or publish, that in whole or in
part contains or is derived from the Program or any part thereof, to be
licensed as a whole at no charge to all third parties under the terms of
The GPL is thus "viral" in the sense that any program making use of code
covered by the GPL becomes code covered by the GPL. The GPL has not yet been
tested in court. Thus precisely what "derived from the Program" or "based on
the Program" (another term used in the license) mean is not settled.
The GPL is the source of considerable controversy among open-source software
producers and their proprietary counterparts. Some open-source advocates
believe that the GPL retards the growth of open-source software by
discouraging the development of proprietary software that can work with open
source and by limiting the profit opportunities of firms, such as Caldera
and Red Hat, that want to promote open-source products. Proprietary
firms-Microsoft has been the most vocal-have expressed concern that they
risk losing their intellectual property rights if they make their
proprietary software interoperate with software covered by the GPL.
Although the FSF has been quite influential, the open-source community does
not speak with one voice. The Open Source Initiative was founded right after
Netscape announced in early 1998 that it was going to release the source
code for its Web browser. "We realized that the Netscape announcement had
created a precious window of time within which we might finally be able to
get the corporate world to listen to what we have to teach about the
superiority of an open development process," explained Eric Raymond-that "it
was time to dump the confrontational attitude that has been associated with
'free software' in the past and sell the idea strictly on the same
pragmatic, business-case grounds that motivated Netscape."
To this end, the OSI has published a guide to the terms a software license
should (and should not) contain in order to be considered open source, as
well as a list of licenses that satisfy its definition of open source. In
particular, it permits the removal of the "viral" aspect of the GPL by not
requiring the redistribution of a modified program under the same license
terms as the original.
One of the non-viral licenses that the OSI recognizes as open source is the
Berkeley Software Distribution (BSD) license. A user who chooses to
redistribute a BSD-licen ed program (whether modified or not) must retain
the copyright notices in the program code, but faces no other serious
restrictions. As a result, anyone can modify a program subject to the BSD
license and distribute the modified version as proprietary software, with no
obligation to make the modified source code available to other users.
The latest operating system for the Apple Macintosh, OS X, uses elements of
a BSD version of the Unix operating system. The Apache license also allows
redistribution without major restrictions on the user. These non-restrictive
licenses were relatively popular in the past, but whether they will continue
to prove appealing to the people who start open-source projects remains to
The Economics of Open Source
To date, open source has mainly involved programmers working on projects for
IT professionals and other programmers. Open-source programs aimed at
sophisticated users, particularly server operating systems and server
applications, have done quite well: witness Linux, Apache, and Sendmail.
These programs have proved popular because they meet techie needs. The fact
that they are free and the user gets the source code helps, too. Open source
has fallen short for average PC users, though, who demand software that is
polished and easy to use. Non-techies would rather pay for something they
can install right out of the box, and access to the source code does not do
them much good.
The economics underlying open source sheds some light on whether this is a
temporary state of affairs, as some believe, or whether open source will
remain a techie-only success. I start by considering why individuals donate
their time to open-source software projects and then examine the investment
by for-profit firms in the development and distribution of free software.
There has been considerable debate about programmers' motives for working on
open-source software projects without compensation. Joshua Lerner of Harvard
and Jean Tirole of the University of Toulouse have provided the most
First, programmers may invest time in open source because the immediate
benefits they can capture in their own work outweigh the costs. As long as
they generate some net benefit for themselves, they have incentives to do
the work and no particular reason not to share it with the larger community.
For example, the people who initially developed Apache were information
system administrators who had a specific set of problems that needed to be
solved and had no proprietary alternative.
Second, programmers' success in solving a problem and making a contribution
to the larger effort brings them respect among their peers. While economists
tend to underestimate such intangibles, I believe this is an important form
of compensation in most professions. In open source, contributions are made
public through credits posted in the software code. Raymond has observed
that "surreptitiously filing someone's name off a project is, in cultural
context, one of the ultimate crimes." Successful contributors often obtain
professionally prestigious positions in the project administration.
Third, that same success may signal their abilities to the job market and
help them get lucrative work. Many of the Apache developers have been hired
into proprietary software development groups.
Fourth, association with a successful open-source project may also help them
lure investors. Marc Andreessen of Netscape attracted venture capital-and
venture capitalists in the person of James Clark-thanks to his success in
developing the Mosaic browser, a project with widely available source code
that in its day was the most downloaded software program on the Internet.
It is too early to tell whether investments of capital in open source will
yield the same success that individual investments of time and expertise
have generated. Most open-source business strategies have involved using
open-source software to sell complementary products or services. Such
businesses typically fund programmers who participate in open-source
projects, with the rationale varying from a need to keep abreast of the
technology, to solving problems that are important to their businesses, to
increasing the supply of open-source software generally.
Caldera and Red Hat, for example, generate revenues by charging for bundles
of open-source software that include Linux, as well as by providing user
manuals and ongoing customer support. Both firms are also implementing plans
to charge for updating computers with the latest Linux and related
IBM, an enormous presence in proprietary software, has funded the
development of open-source software that it can distribute with its
proprietary server hardware. It found Linux appealing because it sells many
different types of network servers, using incompatible operating systems.
Running Linux on all these servers would permit IBM to unify its server
hardware lines, from small servers (basically PCs) to mainframes. IBM
presumably expects to recover its investment from sales of hardware and
proprietary software. So long as the open-source products the company
develops primarily increases the appeal of IBM hardware rather than that of
its competitors, it does not face any problem from the fact that the source
code is available for free to others.
It remains to be seen how much for-profit enterprises will invest in
open-source projects. All have strong incentives to free-ride on the
open-source development efforts of others unless they can figure out a way
to capture the returns on investment through the sale of complementary
products. And that requires some market power over the sales of those
complementary products, so that they can charge enough of a premium to earn
the investment back.
IBM may have incentives to invest in open-source projects to the extent that
it makes IBM hardware more valuable to customers. But it is hard to see how
Red Hat and Caldera could charge a significant premium for bundling and
support services, since there are no barriers that would prevent rivals from
entering these businesses and competing away profits. Some other firms with
business plans based on open source have already encountered that problem.
Eazel developed an open source program to manage files and hoped to make
money (as do Red Hat and Caldera) through automated software downloads. It
is now defunct. VA Linux Systems was a famous early firm that sold server
computers with Linux installed. After discovering that it could not compete
with the likes of Dell and Compaq, it has left the hardware business
Open-Source vs. Proprietary Software
Open-source and proprietary software production methods have strengths and
weaknesses that vary according to the type of software and the
sophistication of the users.
The proprietary software model offers two key advantages. First, the ability
to capture profits by restricting access to the fruits of their labor
encourages software firms to make risky investments in the development and
improvement of software. Creating a state-of-the-art computer game such as
Riven (the successor to Myst) can take years and cost millions of dollars.
But the importance of the profit incentive goes beyond the creation of a
particular software package. Most of the major software companies engage in
research and development that may eventually spill over into how software is
Second, by owning the intellectual property rights, a for-profit firm can
control the development and subsequent evolution of the product. While it
may make mistakes in shepherding the project, it has profit motives to make
the program as valuable as possible to consumers. Most mass-market software
vendors, for example, have chosen to make sure that their products are
compatible with previous versions, and to make sure that there is a
"standard" current version. Having standard versions encourages the
development of complements-for example, the vast array of applications
software for Windows.
The proprietary model has some weaknesses, though. First, to make profits
proprietary firms must limit access to the source code sufficiently to
prevent copying. Most vendors of proprietary mass-market software choose not
to make the source code available, even to customers. That makes it
virtually impossible for customers to fix bugs they encounter or to
customize software to their particular needs. This, of course, is mainly an
issue for proprietary software used by information technology professionals
who have the ability to modify source code. It is not an issue for most
consumers, who have no desire to get under the hood. But when the issue
becomes sufficiently important to large customers, proprietary software
firms have been known to give these customers at least limited access to
their source code.
There is a second weakness. Although we generally think of free market firms
as being quick to capitalize on profit opportunities, in the fast-paced
world of information technology some customers may have needs that are not
being met by software firms. Again, this is an issue primarily for IT
professionals, and it is not clear that proprietary firms have failed to
deliver the bacon very often.
The open-source method, for its part, has several important strengths. The
first arises from the nature of intellectual property. Once created
(obviously an important pre-condition), society benefits the most when
intellectual property is distributed at marginal cost-which, in this case,
means for free. With most open-source software, no one has to pay for the
software (in monetary terms at least-ease of use and lack of warranties are
other matters) and everyone can use the intellectual property as they see
By the same token, the fact that the source code is freely available
maximizes the value of the intellectual property by making it possible for
users to tailor software to their own needs. Last, by distributing the
source code and providing a clearinghouse for reviewing fixes and extensions
to the source code, open source provides a way for a large number of users
to help with debugging and extending the software at minimal cost.
But the open-source method has disadvantages as well. First, participants in
open-source projects have limited opportunities to realize a return on their
investment of time and effort. Of course, they may have non-financial
motives (enjoyment of programming, respect of peers). But there is no
particular reason to believe that they are any more powerful in software
than in other products, where the proprietary model has dominated. Firms
inclined to hire programmers to work on open-source software also seem to
have limited profit potential, except in special circumstances. All else
equal, open-source software appears likely to attract less effort-purely
creative as well as nuts-and-bolts programming-than proprietary software.
There is a second disadvantage. There does not appear to be any method for
the open-source software method to identify and respond to the needs of
non-techies. Part of the problem is that the inability to earn profits
limits the opportunity to conduct serious market research. The other part is
that programmers will probably invest less time in working on software from
which they gain no direct benefits.
Moreover, because open-source software tends to be written for use by IT
professionals, there are few incentives to smooth out the rough edges and
make software easier to use . And that limits the prospects for open source
in the software used by the vast majority of consumers. In 1999, it's worth
noting, software not used by IT professionals generated a little less than
half of all revenues from packaged software sales.
The Future of Open Source
Although experience suggests that surprises are possible, open source does
not seem a viable model for producing mass-market software. To see the
difficulties, consider Linux. In many ways Linux seems well positioned to
become an operating system for "client" computers-the hundreds of millions
of desktop and laptop machines at the end of networks. It is used by IT
professionals on millions of network servers and has a number of good
computing properties-for example, a disinclination to crash. But it has made
only limited progress in evolving towards an operating system that companies
could ask their employees to use, or that home users would want to use.
Linux doesn't have a standard, easy-to-use graphical interface, although
work in that area is continuing. And it can't boast of many high-quality,
user-friendly applications that appeal to mass-market users. It also seems
to offer fewer standard hooks for applications software developers to use in
developing complementary programs with a consistent look and feel.
No doubt, Linux will overcome some of these weaknesses. Nevertheless, it is
hard to imagine what financial or non-financial incentives would make
programmers spend time working on products such as spreadsheets or
statistical software that they do not use, or why they would be particularly
good at creating high-quality software in areas far from their interests or
Open source is more promising for software that is written by techies for
techies. However, several factors may retard the spread of open source. Many
software products-and especially operating systems-have become popular
because so many other software products work with them. The Free Software
Foundation's GPL, however, makes it difficult for proprietary firms to
produce software that complements open-source efforts because it is too easy
for the proprietary code to catch the GPL virus.
Another problem is that producers of complementary software and hardware
products tend to exchange technical information during development cycles.
The decentralized evolutionary model used in open source may make such
coordination much more difficult.
The thousands of programmers around the world who volunteer their time for
open-source projects provide the movement's greatest strengths. But one
wonders how much effort these volunteers will offer to projects that are
important to the future success of open-source software, but are neither
intellectually challenging nor of immediate utility to themselves.
Likewise, open source could become a victim of its own success. As it
evolves from a hacker phenomenon to an industry backbone with formal
institutions, it is easy to imagine that those who offer assistance because
they like being part of a cool club will find it less alluring-sort of like
the difference between hanging out at a funky café and one of the six
Starbucks on your block. And if companies really do start making significant
profits by producing services that feed off the open-source movement, one
can also imagine that unpaid volunteers will lose their enthusiasm for
Las discusiones por Internet son como las Olimpiadas especiales.aunque ganes
todavía sigue siendo retardado.
Las discusiones por Internet son como las Olimpiadas especiales.aunque ganes
todavía sigue siendo retardado.