The Liberal Source Essay

Version 1.1
Original published in 1999 on lss.com
Original Author: Paul johnson

This essay written in 1999 underlines some interesting points that were partially reused in the new Sustainable Software Definition. However this text should be treated as a global explanation of liberal source systems and not as a definition of sustainable software. Please refer to the sustainable software definition to know the key rules underlying sustainable software.

There are two worlds in software today: the closed source world of traditional software vendors, and the open source world of Linux and the Internet. Neither the open source nor closed source models are new: both have been around almost from the beginning of computing. But neither has managed to kill off the other.

In this essay I want to look at both models of software development and then propose a new model which combines the best of both.

Capitalism

Before going on to look at open source software it is worth looking at the reasons why closed source software has survived so well up until now. To do that we first need to look at capitalism and the market economy.

Capitalism has managed to out-perform every other model of economics which has ever been tried. There are a number of reasons for this, some of which are listed below:

It is self-optimizing. Every other system requires a big, expensive bureaucracy to organize economic activity. By contrast capitalism employs the self-interest of actors in the marketplace. Decisions about the allocation of resources are made in an optimal manner because anything else would be to the detriment of the people making the decisions.
It rewards initiative and innovation. Risk-takers are rewarded in proportion to the risks they take. In contrast bureaucrats tend to take safe decisions and thereby miss opportunities.
It rewards people who act for the benefit of society, broadly defined as the greatest good of the greatest number. The butcher, the baker, and the candlestick maker do not carry on their trades from charity or a selfless desire to serve society, but from the fact that they get paid for doing so.

Capitalism is often criticized because it does not always meet these goals. For example there is a tendency for successful companies to buy out or bankrupt less successful ones, eventually leading to a monopoly of supply. Such "market failures" call for government control. This essay is not going to go enter the debate about the precise circumstances under which such control should be exercised, but in general it should only occur when the market would create a greater social cost than the government intervention.

Comparative Advantage

The theory of comparative advantage is possibly the only part of Economics that is both non-trivial and certainly true.

First consider a theoretical example. There are two countries, A and B. They both produce two commodities, cotton (C) and steel (S). A is capable of making 100 tons of steel or 100 tons of cotton, or any combination of the two. In other words the output of A is

C+S = 100

Country B on the other hand is much poorer. It can only make 50 tons of cotton or 25 tons of steel, or any combination of the two. Its output is

C + 2S = 50

Now let them trade. Suppose that A gives B 20 tons of steel in return for 30 tons of cotton. A now has 10 tons more goods than it had before. To put it another way, its total wealth can be shared between cotton and steel as follows:

C+S = 110

This means that A is clearly better off than before the trade.

B has less tonnage than it had before, but the steel it now has is more valuable to B than the cotton it traded. Its equation now looks like this

C+2S=60

This is the magic of trade: it makes both sides better off than before. It also means that B is better off concentrating its efforts on producing cotton because it can trade the cotton with A and get more steel that way than it could produce by itself. Conversely A will be better off if it moves its resources into steel production and trades with B to get the cotton it needs. In the terminology of economics, A has a comparative advantage in making steel, while B has a comparative advantage in cotton. Note that this is completely irrelevant to their absolute positions: if A and B have the same populations then the GNP per head of B is half that of A, and wages in B will be correspondingly low. But the trade still benefits both countries.

The really neat thing is that if both sides aim to maximize their own profits on the trading, the two states A and B together will produce the optimum amounts of cotton and steel for both their requirements.

This works on a personal level as well. I can choose to iron my own shirts, which takes me two hours a week, or I can employ a laundry to do it for me. It takes the laundry one and a half hours because they are better at ironing than I am, and they charge me £15. I can earn £20/hour programming computers, so I am better off spending another two hours at work and hiring a laundry. I have a comparative advantage in programming computers, and they have a comparative advantage in ironing shirts.

Sir Winston Churchill was an excellent amateur bricklayer, better than many professionals. Despite this he still hired bricklayers because he was even better at politics than he was at bricklaying. Bricklaying was just a hobby.

Trading Risk

Risk is an inevitable part of all our lives. However any given disaster can present a bigger problem to some people than to others. Therefore it makes sense to shift risk over to the broadest shoulders, and capitalist economies have developed a wide range of ways of doing this.

The simplest risk trade is insurance. If my house burns down then I will suffer great monetary loss and in addition I will be homeless and unable to afford further housing until my mortgage is paid off. This risk is unacceptable to me, so I pay an insurance company to take it off my hands. In return for an insurance premium they will shoulder the financial costs of a fire.

An employment contract is another form of risk trading. The amount of work in my profession may vary. I could spend months without work and then be faced with a glut. My employer accepts some of this risk as part of my employment contract: as long as they carry on employing me I will receive a constant wage whether there is work or not. When work is slack they can afford to keep me idle because when there is work I can earn enough to pay for my wages during the slack periods.

The most sophisticated risk trading is done by traders in the world of financial derivatives. A company needing to buy or sell a particular metal can reduce its exposure to the risk of future price fluctuations by setting up contracts for futures and options in the metal in question.

The big problem for many risk-trading schemes is that of "moral hazard". This occurs when someone who has payed someone to take a risk away is then less motivated to avoid the disaster in question. For example, having purchased theft insurance for my car I might be less scrupulous about parking in a well-lit area. In the most extreme cases moral hazard shades into outright fraud.

Software Capitalism

At first sight the operation of capitalism on the market for software appears straightforward. Capital is invested in the production of software, of which the largest chunk is wages for software engineers. As a result of this labour a software package is produced and copies are then sold to recoup the cost of production and make a profit.

However there is a problem. Once the investment in the software has been made the copies are virtually free. Anyone else could therefore produce copies without paying the original authors. This would prevent the authors from recouping the cost of production. This is an example of market failure, and the solution is one of the earliest forms of government intervention in the free market. It is the body of law known as "intellectual property", or IP. Under this the originators of a software package have the right to determine who is allowed to make copies.

The pricing of software follows a similar principle to the pricing of any commodity, and is represented by the following graph:

This shows a graph of demand against price. The line represents the price at which different numbers of people are prepared to buy the software. As the price decreases so more people are willing to buy the software and so the sales increase (a very few luxury goods have a positive slope: increasing the price increases the demand). The threshold price at which a particular individual will decide to buy is a measure of the value of the software to that individual: any higher and its not worth buying. On the other hand if the software can be purchased for less than the threshold then that represents a net benefit for the purchaser: in exchange for, say, £100 the purchaser gains software worth, say, £120 to him.

Its usual in graphs of this sort to have another curve going from bottom left to top right, representing the amount that will be supplied at different prices. The "equilibrium price" is the point where the two curves intersect. Its easy to show that any attempt to move the price away from the equilibrium is counter-productive: you either get a shortage and a black market, or you get oversupply and have to throw away the surplus. But a software package doesn't have a supply curve: once the first copy has been produced the subsequent copies are effectively free.

Areas on this graph represent value. The value of the sales to the software company is area B: the price charged multiplied by the number of people who buy. Area A is the excess value to the purchasers. Both these areas represent a benefit. The software company is paid and the purchasers are better off. But area C is a "dead-weight loss" to everyone. This represents the people who could make use of the software, but not enough to justify the purchase price. Ideally we would like to eliminate area C, but that can only be done by giving the software away. The software company meanwhile does not care about area C. They just want to maximize area B: their revenue.

Fortunately the software companies are aware of this problem, and have hit on a partial solution. Instead of charging a single price, they divide people up according to where they appear on this chart and charge different prices. Area A (green) is much smaller, and area B (blue) is much larger. But area C (red) is what counts, and that is much smaller. The loss of value to the society as a whole has been greatly reduced. If the companies can identify those who are in area C then they can even extend the discount to zero, allowing everyone to benefit from the software and eliminating area C completely. From the global point of view it doesn't matter whether the value of the software is given to the software company or the customers, provided that someone gets it. So this kind of discount scheme is a Good Thing.

Of course in reality its nowhere near that simple. The consumers don't want to pay more than they have to, and the software company is therefore restricted to categories it can easily verify. And of course nobody really knows exactly what shape the curve is. Also there is a feedback effect in software sales, whereby the increased popularity of a package for, say, word processing makes it more valuable to potential purchasers because they can exchange documents with more people. Managing the resulting monopoly is an ongoing problem and one that I will return to later in this essay.

The Inefficiencies of Commercial Software

One would expect that the developers of commercial software would have concentrated on efficiency in order to stay competitive with the rest of the market. However this frequently does not happen. Large organizations do not run internally on the laws of capitalist supply and demand. Instead they resemble more closely the command economies of communism.

Success in such a system is not gained by doing what is best for the company or most efficient for the economy, it is gained by from ensuring that those above you in the hierarchy believe you to be doing what is best for the company, or at least for their part of it. Since they have less than perfect information about you, whatever method they use for assessing your behavior is susceptible to manipulation.
A large organization will contain sub-cultures in various departments, and members of these sub-cultures will tend to seek status within their sub-cultures as well as in the formal company hierarchy. These status-seeking behaviors can often be destructive to the company.
Artifacts are valued by measuring the cost of the inputs rather than the extent to which they reduce the cost of other activities in the organization. Wasted work and inefficient practices are thereby hidden from those in a position to correct the problem.

The resulting mix of cynicism, petty corruption and perverse incentives is brilliantly satirized by Scott Adams in his "Dilbert" cartoon strips. All large companies have to deal with these problems, and in fact this is one of the main tasks of the senior management.

Free Software

Some writers (notably Richard Stallman) have chosen to see the red area in the graphs above as a demonstration that the capitalist model of software production is fundamentally immoral and inefficient. If charging for software reduces the value of that software to society as a whole then the only moral course is to give software away for free. They also point out the loss of efficiency caused by the hoarding of knowledge. This is the origin of the Free Software movement, now partially rebadged as the Open Source movement.

A great deal has been written recently about the advantages of open source software. Some writers have predicted that it will overthrow the commercial software world. In its present state I find this improbable. Open source software has too many inherent limitations. These are discussed below. However the strengths of free software are very real, and I will return to these later on.

There is, of course, almost no such thing as free software. If a programmer writes some software then it is at some cost to somebody, even if that cost is only the payment that the programmer could have had for writing something commercial (known to economists as an "opportunity cost"). The only exception is software which is written as a recreation. This provides value to the programmer (he gets to do something for free which he would have been prepared to pay to do) and also to society (in the form of useful software).

Hang on a minute. If this is so, why do so many people write such good software for free?

Some of the incremental progress in free software can be attributed to the value of existing free software. If someone finds free software that almost fits their requirements then a few extra features can often be added for less than the cost of switching to a commercial package which could handle them.
Some software is written by government employees, and is therefore automatically placed in the public domain. The GNU Ada Translator (GNAT) is an example, as is the WinVN newsreader.
Increasingly, software companies are releasing the source code to their products in the hope that other programmers will extend them for free.

But these things alone cannot explain the great mass of high quality free software which is available today. The only way to explain it is to look at values that cannot easily be quantified in financial terms. And to do this we need to look at evolution and anthropology.

The Evolutionary Biology of Free Software

[...] her clients had money, which was useful, but they also paid in respect, and that was a rock-hard currency.
Equal Rites, by Terry Pratchett.

Trade is one of the oldest human institutions. Archeologists have found stone tools which can be traced to ancient quarries hundreds of miles from where the tools were found. More recently anthropologists have made contact with tribes that had never seen a white man before, but were still able to obtain mass-produced metal knives and axe heads. They got there through trade with neigbouring tribes. Possibly the most amazing example is the traces of cocaine and nicotine found in Egyptian mummies. Coca and tobacco only grew in America, half a world away. To get to Egypt they must have been traded around the Pacific rim and across Asia, increasing in value exponentially as they went, until by the time they reached Egypt they were almost priceless. Almost, but not quite.

Barter is the simplest form of trade. I have a knife that you want. I am hungry, and you have a freshly killed wild pig. Lets swap. But barter is a very inefficient form of trade. It requires that two people have something that the other wants in the same place at the same time. Suppose I'm hungry now, but I don't have anything to trade for the pig. Can I promise you that, if you share your dinner tonight, sometime in the future when our situations are reversed I'll share then? Supposing you agree. How can you trust me? But if we can't work out a deal then I'm going to be hungry tonight and you are going to be hungry at some point in the future.

In the modern world we solve this problem with money. Money acts as a store of value. Even in cultures without money there is often a commodity which serves the same role (examples include particular kinds of sea shells, gold, amber, cigarettes and illegal drugs). But it turns out that money is not actually necessary to this process. Instead we do it with respect.

This section opened with a quotation from "Equal Rites" by Terry Pratchett. The subject of the sentence is a witch, and Pratchett's witches live on respect. Most are completely outside the cash economy. If they want something then they only have to ask for it. The mixture of gratitude and fear in the rest of the population often means that they don't even have to ask.

Anthropologists have observed respect working in the real world as well. In hunter-gatherer societies even the best hunters will only bring meat back occasionally. When they do they usually have far more than they can cope with at once. The obvious solution is to share. But this brings a problem: what about the lazy hunter who never brings anything back but still gets a share of everyone elses hard work?

Respect is the solution to this problem. Every time a hunter brings home the bacon he goes up in the estimation of the rest of the tribe. This translates into choicer cuts of all the meat, not just his own. In effect the hunter has traded meat today for meat tomorrow, using the respect of others as a store of value on which to draw during lean times. And somewhere along the line respect is usually traded for sex and reproduction as well. This last point is important, because to understand why we naked apes behave the way we do it is necessary to understand the way in which evolution has shaped us.

An Aside: The Peacock's Tail

Why do peacocks have such huge tails? Surely a bird with such a tail is less fitted to survive than one without. Natural selection should weed out the peacocks who spend valuable energy on their tails, favoring instead the ones who spend it on bone and muscle. But peacocks, not having read The Origin of Species, continue to flaunt their tails.

The answer to the riddle is in a curious feedback loop between peacocks and peahens. The hen wants to choose the best possible father for her chicks. Initially that meant picking the healthiest, strongest peacock. Feather condition is a good indication of general health, so peahens evolved to pick the peacocks with longer, glossier feathers.

(I'm using the word "want" as a shorthand. To be more precise, there is selective pressure favoring peahens who make this choice. But that is a very clumsy way of saying the same thing. Just remember that "want" implies selective pressure rather than conscious choice.)

But this means that the peahens will favour peacocks who put all their energy into their plumage. Plumage started out as a simple matter of survival, but now it has become the road to reproduction as well. Better to be a weak peacock with brilliant feathers than a strong peacock with unimpressive feathers.

One might expect that, since the male's feathers have become mere advertising with no real relationship to the health of the bird underneath, the peahens would evolve to use another criterion for picking mates. But this is where the feedback becomes important: a peahen picking a mate doesn't merely want strong chicks, she wants chicks who are going to reproduce when they grow up. That means producing male offspring that meet with the approval of other peahens, and the best way of doing that is to pick a male who already meets with that approval. If a peahen tries to buck this trend by picking a male with a small tail then she is going to produce small-tailed male offspring who will not find any mates. And so the peahens and peacocks are locked into a cycle of producing ever more useless and showy tails.

Back to the Plot: Selecting for Respectability

I ended the paragraph before the aside about peacocks by noting that respect is often traded for sex. The reason for this is very similar to the story of the peacocks: if you were a prehistoric female then picking a good hunter to father your children meant that they would, on average, be good hunters as well. (It also meant that this particular good hunter had a direct interest in looking after your particular children, but that is another story). Prowess in hunting brought respect, so a cave-woman would want the most respected caveman around. But if all the other females make the same decisions then being being respected takes on the same importance for humans as having a huge tail does for a peacock. It ceases to be means to an end, and becomes part of a cycle of increasingly inefficient behavior. This is the origin of such strange behaviors as Victorian Respectability and gift cultures. And it is gift cultures that interest us here.

In a gift culture the practice of sharing property to gain respect can go far beyond any concern about future hunger. Lavish parties are thrown, with expensive gifts for all the guests. In this way the giver gains respect in the community, and ultimately reproductive success. In extreme cases the property is not even shared, it is simply destroyed. In the tradition of "potlatch" expensive skins would be drenched in expensive oils and then set on fire as an ostentatious show of wealth (whilst other men pretended to be freezing cold in order to minimize the credit due to the host) . A potlatch is the behavioral equivalent of the peacock's tail: its completely useless to all concerned, but you've still got to do it if you expect to breed. And its important to understand that the men engaged in competitive gift giving are doing so out of instinct, not because they have consciously planned it as a route to reproductive success.

The hacker community is another gift culture. Its inhabitants gain respect not by amassing property but by giving it away. And the word "community" is important as well. It means a particular sub-culture. Humans evolved to cope with a community of a few hundred people at most, not hundreds of millions. We solve this problem by creating mini-communities and seeking respect within those. Hackerdom is a cluster of these mini-communities and hackers seek respect within them by contributing code "for free".

The Economics of Respect

So far we have seen how "respect" acts as a form of value. The analogy with money is quite close. Like money the supply of respect is limited, and it is valued by members of the community. However there is one big difference: respect cannot be traded between communities. I might gain respect amongst the community of hackers, but that won't get me free credit from a shopkeeper, even though he might be happy to extend it to his local priest. And most of hackerdom would look in askance at the priest if he wanted a favour from them. The two communities are separate, and respect in one does not buy respect in another.

In contrast money is freely exchangeable between communities. If I want something from a shop I have only to walk in with money. It matters not to the shopkeeper where I got it from. This is the great limitation of gift cultures: they can only barter with the outside world. To do more you need money.

This limitation means that money has a divisive effect on gift cultures. A member of a gift culture is tied to that culture: the store of respect he has earned is in the heads of other members of the culture and cannot be transferred elsewhere. But if the same individual obtains money instead of respect then that money can be taken elsewhere, and even traded for respect in another gift culture if desired.

The lack of mobility in respect matters in another important way. This essay started by discussing the advantages of capitalism. One of these was that capitalism is self-optimizing. People in the capitalist system make decisions to maximize their return and minimize their costs. Overall these decisions are usually the most efficient ones possible. The two countries in the discussion of comparative advantage are a good example. But when you have stores of value which cannot be traded then these decisions cannot be made in an optimal way. This lies behind the greatest failure of the free software effort: until recently its products could only be used by other hackers. Documentation was sparse and poorly maintained, and user interfaces were arcane and forbidding. Even today there is no free software office package that comes close to the flexibility and ease of use of the commercial packages. The reason is simple: hackers don't want easy-to-use office packages, and so would give little respect to to anyone who wrote one. If you want the respect of hackers then you are much better advised to write something that they want. The rest of the world needs easy-to-use software to handle routine administrative jobs, but the money behind that need does not buy respect in the free software community.

Some hackers do manage to parley their respect in the hacker community into money. Some gain good jobs on the strength of reputations built amongst hackers. Others run consultancies, publish books or provide support. Promoters of free software point to these examples as evidence that capitalists and hackers can work together, but the basic inefficiencies remain. There is also considerable inequity in such an arrangement. The consultant or publisher makes money from free software, but all the hundreds of programmers who helped write it get nothing.

Liberal Source Software: A Synthesis

So far this essay has described the motivation behind the creation of both open and closed source software. Closed source software is driven by the economics of money, and open source software is driven by the "economics" of respect and status.

This section takes a new tack: it suggests a synthesis of the best points from both the open and closed source software worlds. We need to keep both the self-optimizing nature of capitalism and the openness of the respect-based communities of open-source development. This synthesis is Liberal Source Software, or LSS for short. In an earlier version of this essay it was named "Commercial Open Source Software" (COSS). However it does not quite fit the definition of "Open Source" software provided by the Open Source Initiative.

The great advantage of open source software is the "many eyeballs" effect. The open source community can bring immense resources to bear on a problem, and hackers are not afraid to rewrite code that lacks quality. In the commercial world the pressure is always to release a product with more features, and if code needs to be rewritten then someone is going to be blamed for not doing it right first time.

Making open-source software commercial requires two things: first that payment is collected in proportion to the benefit gained from the software, and secondly that it is distributed to the developers in proportion to their contribution.

Execution-Based Payment

Closed-source software is often referred to as the "crown jewels". Software companies have always assumed that if anyone has access to their source code then they will be unable to charge any money for it.

A little thought strongly suggests that this is wrong. Executables are easy to copy, and the "warez d00dz" can always be relied upon to crack any kind of copy protection or registration key system (incidentally, the warez d00dz are a gift culture where respect is earned by giving away cracked packages instead of newly written software). Few indeed are the people so honest that they have never accepted an illegal copy of a software package from a friend. And yet despite this the software companies still make money.

There are a number of reasons for this:

Software is most valuable when it is used by companies in the course of their business. Managers who pay for licenses for commercial software take the money out of their budgets. Managers who practice or condone software piracy can personally be convicted of criminal offences. This provides a large incentive for honest behavior. However this does break down in the case of small companies where short term cash flow takes precedence over longer term risks.
Disaffected employees who see their company ripping someone off are apt to blow the whistle.
People tend to take the easy course of action. Purchasing a software package complete with manuals and official media is easy, if expensive. Finding a friend who will make an unlicensed copy, and then reading the help file on screen, is usually harder.

This suggests that the authors of open source software could impose licensing charges and expect to have them paid, at least enough to make it worth while.

The source code license for LSS would of course remain free. There should be no charge (other than the usual copying expenses) for obtaining, reading, modifying or compiling the source code. It will also be permitted to execute the software for experimental purposes, for example to determine if it is suitable for the job, or to test modified copies.

Instead, the charges will be imposed for execution which adds value. In a commercial setting this is fairly easy to define: the use of the software to assist in any way with the operation of the company is adding value, and therefore should incur a charge. In a domestic setting the value is defined by the purpose of a particular execution: running a game program provides entertainment, which is valuable, and using a word processor to write a letter saves time and increases the quality of the result.

There are many possible frameworks for charging, most of which have also been tried by the closed-source software companies. For example:

Per-copy charges. A fee is charged for each computer with a copy of the software installed.
Seat charges. A fee is charged for each user authorized to run the software.
Concurrent user charges. The software can be used by anyone anywhere, but may only be used by a limited number of people at any one time.
Per-invocation charges. A fee is charged for each time the software is used.

In the case of the more complex systems some form of automatic logging and accounting would be needed. But that is what we have computers for.

All of these models have advantages and disadvantages, and the best scheme will depend on the package in question. The key requirement would be that the cost to the user would be in proportion to the benefit derived. The aim must be to maximize the revenue from the software whilst eliminating the dead-weight loss.

For many packages it would be perfectly possible to make the lowest license charge zero, provided that no commercial use was made. However the benefit derived from software is not always financial. Game software, for instance, provides no commercial benefit to anybody when it is executed. However it is still perfectly reasonable to require payment in return for the entertainment.

Revenue Distribution

The money from the license fees for LSS will be paid into an account associated with the software package. From there it will be distributed to the developers in proportion to their contribution.

A number of companies have released the source code to their products in the hope of gaining the benefits of free software quality and development, and some have continued to charge for execution licenses even though the source code was free. But none (as far as I know) have ever proposed to distribute their income to the contributors. Yet this is an obvious requirement. It is one thing to write free software to garner respect amongst fellow hackers. It is quite another to do so in order for somebody else to make money. If money is involved then all contributors need to be paid.

Here we run into a problem: just how do we measure each contribution? With the respect-based system of a gift culture this is not a major problem. Each individual contributes some part of the respect, and the system can cope with disagreements about who made the most important contribution. But this will not work with money. Some definite amount of money has to be paid to each contributor. Contributors can compare their payments, and there will always be some who feel cheated.

(Incidentally, this can also be a problem for respect-based reward: Richard Stallman is reported to be annoyed that Linus Torvalds gets most of the credit for Linux despite the fact that Stallman wrote far more of the software in a typical "Linux" distribution than Torvalds)

Requirements

The requirements on the contribution valuation system are as follows:

It must be fair. A large contribution should be rewarded with a correspondingly large proportion of the revenue, and a small contribution should garner a correspondingly small reward.
It must be transparent. All the players in the game must be able to tell if anyone is cheating. For example, simply declaring hours of work would not be effective: contributors would be able to exaggerate their work level in order to gain more money.
It must be self-optimizing. Again, hours of work fails here because some people are better programmers than others. Someone with a comparative advantage in writing documentation should be able to earn more by doing documentation instead of programming, and vice versa. Hence we want to measure the value added to the project rather than the effort spent (and quite possibly wasted) by someone who is not good at their job.
It must be immune to "gaming". This is the practice of carrying out easy work which still scores highly on whatever measure is being used. For example a simple "churn" metric (count of lines added, deleted or changed) is vulnerable to inflation from formatting changes and programming by "cut and paste".
It must be able to reward contributions made in different ways. For example, maintenance of a project web site, writing documentation, cutting code, fixing bugs, administering the money and writing press statements are all important parts of a large software project.

These requirements are somewhat incompatible. On one hand the requirements for fairness and transparency point to an automatic system. On the other hand the problems of gaming and the different types of contribution suggest a system involving human judgement.

Why Not Government Subsidy?

Much free software is already written under government contract, as noted earlier. It is tempting to suggest that this system should be extended. Perhaps software, or at least some kinds of software, should be treated as a public good with its developers paid from the public purse?

The problem is that taxation is inherently inefficient for exactly the same reason as charging for software. Whatever you tax you raise the price of. This produces a deadweight loss in exactly the same way as charging for software. The details of the argument are too complex to go into here, but any microeconomics textbook will explain the principles. The upshot is that the damage done by raising taxes to support the production of software would be about as great as simply charging for the software. And on top of that you have to add the problem of deciding exactly how much a particular piece of software is worth. No mechanism to make this decision has been invented which is more effective than the free market.

Avoiding Monopoly

Software is often used for communication. Any time you exchange data (in the form of a word-processor document, a spreadsheet or an email message) with another person you are communicating, and a basic requirement of communication is that you both use the same method. If you talk by telephone then the other person has to have a telephone as well. If you send them a floppy disk with a Word document on it then the other person must have both a floppy disk drive and a program capable of reading Word documents.

The value of any communication system increases with the square of the number of people using it. If the fraction of population equipped with a particular system is p then the probability that any two people will be able to communicate through it is p² .

This creates a lock-in effect which keeps competing communications systems out of the market. In the telecommunications industry this is sometimes known as the "Ghostbusters Problem" ("who are you gonna call?"). The same problem applies to software. One reason that Microsoft Word is so popular is that most other people have it. Therefore if you want to share documents with as many other people as possible you should use Word. A similar effect applies to Microsoft Windows: most people buy it because most applications run on it, and most developers target it because that is what most people have.

Open Source Software (whether commercial or not) offers a way out of this bind. The key to communication is not a particular software package, but a particular format for the data being interchanged. Closed-source software companies can hide this format but in an open-source software the format can be reverse-engineered from the source code. Hence in an open-source world it is impossible for a single supplier to create a monopoly based on a proprietary data format. If there is a market for programs that read and write that data format then such programs can be produced by anyone and sold independently of the original inventor of the format. The format is still a common communications system, but the inventor is deprived of the monopoly which would otherwise stem from this situation.

A Possible System: Contribution Points

This sub-section outlines one possible system for revenue distribution. It is based around manual judgement of contributions, but with checks and balances to keep the participants honest. Each time someone makes a contribution to a project the contributor is awarded "points" based on the size of the contribution. Revenue is distributed to contributors in proportion to the number of points that they hold. The whole system is managed by a coordinator. Traditionally open source software projects have been guided by a "benign dictator" (as Eric Raymond terms it), and the commercial model just adds the job of administering the revenue distribution to this role.

Points are allocated according to a predefined plan. A plan will describe different types of contribution, for instance 5 points per bug fix, 1 point per new line of code, ten points per week of project coordination, and so forth. In addition it can contain a detailed work programme with each subsystem allocated a certain number of points. The level of rigor in such a work plan is a matter of taste.

Points are only awarded by the central maintainer for contributions that make it into the application. If three people submit a bug fix or enhancement and only one is chosen then the winner takes it all. The others have not added value to the project and therefore do not get rewarded. On the other hand once points have been awarded they cannot be taken away.

There are several checks and balances built into this system:

A contributor may refuse the points offered. The project is then unable to incorporate the contribution. Because this is open source software the contributor can also offer customers an enhanced version of the "official" package and demand payment for the enhancement. In such a situation the users would have to pay twice, once for the main package and once to the person offering the enhancement. This would obviously be inefficient and a very unusual situation. However it gives contributors a lever when bargaining for points.
In theory a dishonest coordinator could take the money and run. However this would not be very profitable. If revenue were not regularly distributed to the contributors then users could be informed and asked to send the money to a new coordinator, or to withhold payment until the matter was resolved. Comparatively little money would be held by the coordinator at any time.
Licensed users of the software would have their payments published along with a license serial number. This would allow, for example, a user or auditor to check if the software has been paid for. It would also allow contributors to see how much revenue has been earned, and hence verify that they are being paid correctly. Companies are increasingly carrying out internal audits to check that all their software is licensed.
The coordinator must keep an account of points allocated and the contributions associated with them. Version control software such as CVS will make this much easier. These accounts must be published, although there is no need to reveal who the points were allocated to. The only requirement to keep the system honest is that all contributors can compare the point allocations for their work as opposed to the work of others.

One potential problem for such a scheme is commercial law. Over the years a substantial body of law and regulation has built up in most countries to cover the practice of buying and selling investments. It is quite possible that a scheme such as this would fall under these laws. However it should usually be possible to avoid their more onerous requirements because contribution points would not be traded. If you are planning on starting up such a project then you should check the rules that apply to you carefully.

Rights of LSS Users and Programmers

The principle behind LSS is that if you derive value from using software then you are obliged to pass some of that value to the authors. The use of LSS software in ways that do not give immediate value remains free.

The origin of the strength of open source software is its freedom (as Richard Stallman puts it, "Think free speech, not free beer"). Users are free to modify open source software and to distribute modified copies. Charges can be levied for the act of copying the software, but not for the information itself. LSS preserves these freedoms. Almost all the requirements for use of the trademarked term "Open Source Software" can be satisfied by a LSS license. The only freedom of conventional open-source software which LSS modifies is the freedom to use the software without charge: Open Source software requires that the software be available for any use without restriction, while LSS requires that certain uses carry a charge.

In the spirit of Open Source Software, modified copies of LSS packages must be permitted, as must the execution of the software for the purpose of testing and demonstrating the modifications. This ensures that the authors of a package do not have a monopoly on improved versions. However the users must still pay the run-time license fees for the original package as well as any license fee imposed for the modification. As noted above, this would be an exceptional course of action, used only to restrict the power of the project coordinator and other authors. It would be in the interests of all the parties to merge the two versions and share the revenue. This would also prevent the authors from creating the upgrade treadmills that conventional software companies use.

Where a LSS library is intended to be incorporated in a larger product the revenue sharing will have to be more explicit. If the larger package is also LSS software then the library authors will want some of the contribution points from it. A closed-source application would require some kind of royalty payment on copies shipped, or possibly an up-front payment.

LSS contribution points would not be tradeable (which is the reason they are not called "shares"). There are advantages to tradable shares: a contributor could obtain money quickly and reduce his exposure to the risk that a project might fail by selling his share of the future revenue in return for immediate cash. This would help to offset the increase in risk that a LSS developer would have to bear from not being an employee of a conventional software company. Unfortunately this also has a number of serious problems.

Once the contributor has sold the share he has no further interest in the success of the project overall. But part of the reason that open-source software communities work is that their members investment of effort is tied to the project. This keeps the interest level up during later phases of the project.
The person purchasing the share may have little or no ability to judge the status of the project. This opens the door to a wide range of investment scams, which in turn are the reason for the many rules surrounding the sale of shares in publicly traded companies.

Distribution Vendors

With the rise of Open Source Software has come the need for pre-packaged bundles of free software. Companies such as Red Hat and SuSe specialize in generating convenient installation disks containing a coherent collection of software (over 450 packages in the latest Red Hat release) ready to be installed on a computer and delivered with a set of manuals. With LSS they can take on the additional role of bundled license vendors. Instead of having to pay hundreds of different software projects for the right to use each piece of software, an end user can pay a single license fee to a distribution vendor who will then distribute the aggregated fees to the various development projects, or even to the individual programmers. These are also the logical people to develop standard end-user license terms and pricing models.

Capitalism and Hackerdom: Collision or Cooperation?

Within hackerdom there are two philosophies justifying free software; the ideological and the economic.

The economic side is principally represented by Eric Raymond. He holds that open source software is good because it is more efficient. If it were less efficient than closed source software then it would represent an immoral waste of resources, and so should not be pursued.
Richard Stallman is firmly on the ideological side. He holds that "software hoarding" is inherently wrong, and that free software is therefore a moral imperative regardless of whether it is efficient in economic terms.

This debate is sterile because it rests on a hypothetical situation: a world in which open source software production is less efficient than closed source production. As many essayists before me have pointed out, the advantage is with open source software and this is not going to change.

The sudden realization of the importance of free software to the world economy has had a mixed reception from both capitalists and free software hackers. Hackers are afraid that capitalists will destroy the gift culture that they have created, and capitalists are unwilling to trust hackers with their future. Neither side really understands or sympathizes with the world-view of the other. The reasons for this impedance mismatch were one of the major motivations for writing this essay.

Hopefully the two sides can find common ground in the principle of giving value in return for value. Trade is the basis of capitalism, and I hope that in this essay I have shown how it is already the basis for open software as well, in the form of respect. By arranging for money to flow instead of respect we can combine the advantages of open-source software with capitalist economics and motivation. The combination should conquer the world.

References

Open Source and Free Software: This essay draws heavily on the ideas variously labeled "Free Software" and "Open Source Software". The first version of this essay neglected to acknowledge this. Like a fish that fails to notice the water until it is taken away, I found that the ideas of Richard Stallman, Eric Raymond and others had become so much a part of my thinking that it never occurred to me to mention them here. See the websites of the Free Software Foundation and the Open-Source Initiative for more details.
Respect and Reciprocity: The pioneering work on this subject, which showed how cooperation could evolve in a population of selfish individualists, is The Evolution of Cooperation by Robert Axelrod. A more recent overview of the subject which covers the economics of respect and gift cultures is The Origins of Virtue by Matt Ridley. This is an excellent book, and the source of many ideas in this essay. It also contains a good explanation of the importance of comparative advantage at every level from individuals to nations.
Status and Status-seeking Behavior: An interesting but now rather dated discussion of this topic is The Human Zoo by Desmond Morris, better known for The Naked Ape. Unfortunately Morris was writing before the development of the theoretical background described in The Origins of Virtue.
Supply and Demand: Any decent micro-economics textbook should cover the topics of supply, demand, taxation and deadweight loss. I happen to have used Microeconomics by Edwin Mansfield. The Economist also publishes occasional briefings on economic topics, which is where I first came across the idea of comparative advantage.

Paul Johnson - 1999