The assertion "Information wants to be free" is a peculiar expression of desire on the part of an abstraction. This essay distinguishes between software and information and argues that a claim that may apply to software is applied unthinkingly to "information".
How to Troubleshoot Apps for the Modern Connected Worker
The Myth of Free Information
1. I began this essay as a letter to the New Yorker, but it got out of control. It is a review of a
review of a book.
The Myth of Free Information
by
Kamesh Aiyer
Malcolm Gladwell’s review of Chris Andersen’s book (New Yorker, July 6/13, 2009)
continues a myth about “free” stuff that has plagued us for many years. The intellectual
foundations of Chris Andersen’s musings lie (if they lie in just one place) in the “Free Software”
movement that Richard Stallman and others initiated thirty years ago with the Free Software
Foundation (FSF). Richard Stallman, a super-programmer in his own right, came to believe that
the knowledge embodied in software should be free, not owned as property, and therefore not
bound by patents or copyrights. Arguably, his efforts have contributed to successes like the
World-Wide Web, Linux, Java, the open source movement, and so on. But, these successes, of a
limited model of what should be free, have led to a myth that all information should be free, that
“information wants to be free”.
The vendors of this concept, like Chris Andersen, point to the four arguments that Gladwell
identifies – Technological (“the digital infrastructure for delivering information is effectively
free”), Psychological (“Consumers love it, preferring free to a small charge”), Procedural (“there
is no such thing as too much information, so the price does not force a judgment call as to what
to buy”) and Commercial (“you can make money by giving it away”). As Gladwell points out,
each one of these arguments fails as we look at the real economics of giving away valuable
information.
We need to be careful about the conclusions we draw from Gladwell’s essay. Just because
software needs to be free does not mean that information needs to be free; contrariwise, just
because information is not free does not mean that software cannot be free. The successes for
2. free software, such as Web browsers, email clients, java, and so on do not create a case for free
information and the failure of free newspapers will not mean that free software will fail sooner or
later.
The assertion that software should be free originated with Richard M Stallman, an
iconoclastic product of MIT’s Computer Science department, where he was a fixture for many
years. In the later seventies, RMS produced and handed out for free an advanced text-editor
called GNU Emacs. GNU Emacs included a programming language “Emacs Lisp” (sometimes
called “mocLisp”) that allowed any programmer to re-do the text editing interface. As a free,
customizable piece of software, GNU Emacs itself followed in the tradition of an earlier version
of Emacs built on a character-oriented text editor Teco, also from MIT. In this chronological
regression we have reached the computer world of the sixties, a time when the most innovative
research products of places like MIT, Stanford, and Carnegie-Mellon (my graduate school) were
the free software of the day. From time-sharing a main-frame to the personal computer, with
programming languages like LISP, text editors like Teco and Emacs, page processors like PUB,
Scribe, and Tex, a host of free software from universities created a world of elite programmers
who saw this as the natural state of things. This free software made possible a lively community
supported by the military-research complex represented by DARPA (the Defense Advanced
Research Projects Agency). Outside, in the real world, IBM was nurturing a community of
programmers who made a living by customizing IBM’s OS/370, cutting IBM out of the loop.
This was another bunch of people who saw free software as a foundation for making a living.
The seventies ended with a bang, heard in hind-sight. Xerox’s Palo Alto Research Center
produced a personal computer with a screen and a mouse, and Xerox let Apple steal that
business; then, Xerox PARC designed the first laser printer and Xerox produced it, and let the
3. Japanese steal that business; IBM created the PC, a personal computer with an open interface and
outsourced operating system that gave away the store, lock-stock-and-barrel – the operating
system to Microsoft, the processor to Intel, and plug-in hardware business to anybody else with
the stomach for competing in the hardware marketplace. DARPA spread itself thin; industry
realized that maybe somebody could make a profit in this business; and, the universities became
just another place where interesting research could be done. Even as industrial research created a
highly competitive, profit-oriented personal computing marketplace, university research was
creating the internet, email and UUNet, client-server computing, the X Window System, and the
Mosaic “web” browser. This collection of free software would wash over the nineties, a tsunami
whose rip-tide we know as the dot-com boom-and-bust.
In the late eighties, I created a “window-sharing” program for a research group at the Digital
Equipment Corporation. A central server enabled a number of personal computers running the X
Window System to share the applications on their screen. Even at that time, this was not an
original idea – others were doing similar things. One of the responses we received when we
demonstrated the product, was that it was a great idea and that we should distribute the client
software for free and only sell the server or, better still, charge a fee for operating the
service. The argument made was not that “software wanted to be free” or some other variant on
the theme of free software, though many of the people who offered this advice came from the
university research world. The argument was an analogy to Gillette’s success with selling razors
– the razors themselves were ridiculously cheap but the blades were proprietary and were the
source of profit.
Gillette was the “reference model” for the free software movement – free razors, expensive
blades. But what was the analogue to the free razor and what was the blade? “Free the
4. infrastructure, pay for maintenance services,” said some. “Free the clients, pay for operational
services,” said others. But isn’t “service” part of the eco-system, just like
“infrastructure”? Aren’t the clients the blade? The analogies did not make for clear thinking.
Something had to be free, to draw the suckers in (razors – “Psychology”), something else had to
be charged for, to pay the pipers (blade-price – “Commerce”), something of value had to be
delivered (blades – “Procedural”) and finally, the whole mess had to stink of “Technology”
(whatever).
RMS added a dash of morality to the commercial instinct behind the Gillette model-myth.
Knowledge was a civil right, it belonged to the whole of humanity and could not, should not, be
hoarded. Any constraints to accessing knowledge were immoral – at best, the caretaker of some
nugget of knowledge could charge for maintaining access or to cover the cost of making a copy
(which could include reasonable labor charges as well as amortized storage fees), but could go
no further. No limits could be placed on what the recipient could know. The person did not have
to pay a fee for knowing something. Use, on the other hand, could be constrained – having the
right to know how to make a nuclear weapon did not give you the right to make one. A patent
for a limited time was a concession to the reality that it was often impossible to allocate a portion
of the cost of developing an idea to the price of the idea. A patent on a product does not bar
someone from making a similar product in the privacy of their home, but does bar them from
profiting by manufacturing it. A societal interest is met by supporting a patent for a limited
time. But an indefinitely extensible copy “right”, that limits the ability to express the idea, even
in the privacy (!) of one’s own code, breaks the assumptions behind this compromise. Such a
copyright is not in the interests of society. RMS goes further to assert that it is not in the interests
of humanity, whatever that might be.
5. The software industry has struggled with the problem of establishing ownership of
software. An algorithm represents knowledge of how some computation or process is to be
done; a program that implements that algorithm is an expression in a language of that algorithm.
Copyright protects expression; as such, it has been used to protect programs. RMS complained
that the big players in the software industry had managed, through legal actions and one-sided
agreements, to establish the principle that Copyright extended to the algorithms underlying the
expression. In addition, the unfamiliarity of examiners in the patent office with algorithms
resulted in patent protection being extended to a large number of common, “obvious” algorithms.
Copyright is viral – any work that can be shown to have some copied element of an original is
infected and could violate copyright. But algorithms are procedural forms of mathematical
knowledge. If an algorithm has an addition as one step (computing “a + b”, say), every
expression of that algorithm will have an addition at that step. If one of those expressions is
copyrighted, it should not make the other forms violations of the copyright. Unfortunately, over
time, judges unfamiliar with computational theory have ruled in favor of copyright seekers. The
viral-ity of copyright, coupled with patent protection, had created a monster that threatened the
free exercise of programming as an intellectual discipline.
An extreme example would be the “malloc” program that first made its appearance in the
Unix freely released by AT&T in the late seventies. Malloc reserves a chunk of contiguous
memory for use by the programmer. As such, it is a necessary function and some variation of
malloc has always existed in all computer languages. Malloc is implemented with at least two
code files in the C programming language, malloc.h and malloc.c. There is nothing magic about
malloc, and it is used here as a prototypical instance of hundreds, if not thousands, of similar
programs. Unix received contributions of code from several university and research institute
6. scientists and engineers. Though these were not created by AT&T and no formal transfer of
exclusive ownership to AT&T was ever made, Bell Labs inserted copyright notices in the source
files for malloc as part of the C compiler and of Unix. In the meantime, these same programs
were made generally available and put to other uses by many programmers. These included
competing versions of Unix such as Berkeley and possibly, Linux. In the nineties, ownership of
Unix was transferred through a series of acquisitions to Novell which is threatening to sue
vendors of Linux for violating its copyrights. The only way that any vendor can establish that
they are NOT violating copyright is to show a line-by-line comparison annotated by a history
that shows that similar lines derive from different version streams. Generally speaking, this is
not possible – neither Novell, nor the vendors have clean historic code-derivation chains going
back twenty or more years; there have been code exchanges in between, many unheralded; and
any history that could be there is easily manipulated. As the saying goes, “We are talking about
programming, not rocket science.”
RMS established two institutions – the “Free Software Foundation”, a non-profit that has
lead the effort to establish the philosophical/political/economic point that software should be free
;and, GNU (the self-referential acronym “GNU’s Not Unix”), a program to develop a stream of
software that would be free of the copyright claims of the industry because all of the software
would have been developed anew, with clearly established change documentation. GNU
software came with the GNU Public License (occasionally called a “CopyLeft”), a copyright
notice developed by FSF that established a viral claim to all software development that was
based on GNU software. The GPL has gone through a couple of iterations and many versions
are extant now ranging from a strict interpretation of “free” to a version that allows for some
proprietary extensions of GNU software.
7. The biggest impact of FSF has been a growing world-wide “free source” community of
programmers who conceive, prototype, produce, test, and maintain an astounding collection of
software. Despite all the doubts and criticism that has been leveled against the free source
community by software over the past twenty years, some of the best software in the world is free
software.
The free software movement has shown that truly, “software wants to be free”. The small
producers have lead the way. It is an important point to note. Free software is free because it is
the best way for the small producer to put a product into the marketplace, get it tested and
evaluated, and collaborate on improving it. Large dominant companies like Microsoft do not
produce free software except when they are under pressure. Consumers do not want free
software if it is buggy, causes crashes, destroy data, and wastes time. The only free software to
achieve wide-spread acceptance is good free software! Otherwise, the only free software is dead
software!
But software is not information. The free software movement does not even hint that it
should cost nothing to take a free program and run it on a customer’s computer system. The free
software may need to be changed to deal with some specialized feature of the host computer or
host environment. The software may need to be configured and some experimentation may be
needed to establish the optimal configuration. Running the program on the host computer takes
electrical energy and that has its cost. An operator or two may be needed to keep the customer’s
computing infrastructure working; communication lines may need to be established and paid for.
Free software is only free to buy; deploying it is not.
Contrari-wise, there is no producer of free information out there. The large-scale vendor of
information wants the user to respond to directed ads; the small vendor of information wants to
8. be paid now; the search engine developer wants to take over your home-page. This is unlike the
producer of free software who may be paid by the users of the software for continued support or
for adding features or for consulting. Once information has been acquired, there are no further
costs involved with acquiring it or maintaining it – unlike software, information does not
degrade. There is information that loses its value – the weather report for today or the next week
loses its value when the day is past, the price of a stock is most valuable when it is the current
quoted price. The seller of this information can ask a premium price while it has value, but the
degraded value is zero or very close to zero. At that point, the information, if it is to have any
future, enters that new life as a Wiki entry. The only payoff to the author of a Wiki entry is
reputational.
The application of the analogy from razor blades to software and then to information does
not hold. It requires a leap of faith and is not supported by any reasoning. If information is to be
free, the providers of information must make their living in some way other than by providing
information. Somebody must pay for it. Information, sadly, does not want to be free.