The opposite of a correct statement is a false statement, but the opposite of a profound truth may well be another profound truth—Niels Bohr

Material products are getting “smarter” in that more and more of the value of a material product is contained in the information it carries, rather than in its material substance. R. Buckminster Fuller called this process “ephemeralization”[1], and it is one means by which the economics of matter—predicated on the conservations of mass and number—are becoming sidelined by the properties of information. Furthermore, both the capital and marginal cost of making products has trended consistently and rapidly down as manufacturing tools become both cheaper and more versatile, so that the capital cost of an object is increasingly not in the capital equipment required to manufacture it, but in the effort required to design it.

Both the capital and marginal cost of making products has trended consistently and rapidly down as manufacturing tools become both cheaper and more versatile

Figure 1: Ephemeralization: How much value is in the silicon and how much in how it is organized? How useful is this board without being able to look up the data on the chips or having drivers and documentation available for it? There are already a wide range of physical “products” that are more information than matter

Software is only the most extreme example, for which both the “Gnu Manifesto”[2] and the “The Cathedral and the Bazaar”[3] claim that the system of copyright monopoly has begun to behave pathologically, and this has been the basis for the free software movement. Raymond’s arguments in “The Magic Cauldron”[4], suggest that this behavior, and the resulting economic success of free-licensing does not truly rely on zero cost of replication, as even software has non-zero replication costs.

Yet, the business model of masquerading an information product as a matter product—by legally controlling the “uncomfortable” property of free replication—has been remarkably successful for a very long time. This suggests that we should consider whether the converse is possible: can matter-based products be masqueraded to act more like information products—eliminating (or hiding) the uncomfortable property of costly replication ? In other words, can we create a bazaar for free-licensed hardware design information and a matter product manufacturing economy which supports it? And can we do it without poisoning the free-design process itself? If so, we might be able to port the high utility, rate of innovation, and low costs found in the free software community to community-based hardware projects—an extremely attractive possibility.

Can we create a bazaar for free-licensed hardware design information and a matter product manufacturing economy which supports it?

What do we want to keep?

The principle reasons for wanting to emulate free software’s “bazaar” development model are those explored in “The Magic Cauldron”, which identifies the important principles of self-selection and self-management, which are active whenever participation in a project is driven by personal interest rather than profit-seeking.

Raymond breaks management duties into: “defining goals”, “monitoring”, “motivating”, “organizing”, and “marshalling resources”, and shows that when self-selection applies, market forces acting on the participants account for the self-managing nature of bazaars, leaving no need for explicit managers: goals are personal and set before contributors even arrive at a project, motivation is for the project to work so that it will suit developers’ own needs, and motivated people don’t need monitoring. Organization is derived as a special case of goal-setting and a lack of personal conflict over competing ideas, which is natural when no gain other than project success is sought by “competing” parties. This situation is likely to persist so long as there is no material gain to motivate contributors, since “interest” will then remain the primary motivator. All of these properties are as applicable to hardware design as to software.

For the final point, “marshalling resources”, however, hardware is going to present special challenges, since hardware design will require the construction of prototypes, experimental apparatus, testing services, and manufacturing expenses. So, the most serious obstacle to overcome is payment for costs without interference with the interest-motivation of developers:

• A means must be found to offset the materials and manufacturing costs associated with free-licensed hardware prototypes, experimental apparatus, and finally, manufacturing cost.[6]

Hardware design will require the construction of prototypes, experimental apparatus, testing services, and manufacturing expenses

Other requirements for success include the maintenance of a healthy effective bazaar size. This can be a problem for the much smaller hardware developer communities, largely due to a lack of exposure to the tools for manufacturing and design, and there are four strategies that can be used to solve this part of the problem:

• Make the bazaar more efficient at attracting, retaining, and collecting contributions from developers. This makes a larger bazaar size, despite a smaller community. Essentially, it must be easier to contribute.[5]
• Make development reduce loss by more efficient and complete archiving, and above all, usefully cataloging archived data so that it can be efficiently retrieved when needed.[7]
• Make more and better collaborative design tools available to the users as free software so that more would-be developers have access to them, and are made aware of the benefits of free-licensing.
• Actively train users on the tools and methods needed to become an effective developer—encourage growth in the community by turning more users and other interested parties into developers. This can only be pursued by active contributions.

Laying down the law

Almost no piece of hardware is totally “free”, because it invariably uses proprietary products at some depth. With software, a hard legal line can be drawn, but hardware introduces a fuzzy area that needs to be understood and controlled. As Figure 2 attempts to show, we have many different layers to consider: Program, Firmware, PAL devices, PCB design, Core-level, Chip-level, and even discrete component level. It’s less of a problem for mechanical designs, but electromechanical components are frequently single-source, and it may therefore be difficult to replace them in the future. All of which argues for the importance of commoditized subsystems to support top-level free designs.

Figure 2: There are many different levels at which a piece of software or hardware technology can be said to be free or non-free, commodity or single-source; patent-encumbered or not. The situation for hardware is far more complicated and less clear-cut than for software, and free-licenses for hardware design will have to take this complexity into account

In general, we want the depth of free design elements to increase, but we have to be able to tolerate considerable transition time, especially if we want to coexist peacefully with a more conventional manufacturing economy. Just as free programs need free libraries, free machines need free electronics and ultimately free components. However, for pragmatic projects, this will always be a design-trade—sometimes it will be expedient to use components that are proprietary to get the job done, and we will be accepting the risk that those designs will become impossible in the future due to supplier monopolies (the use of “commodity” PCs is an example).

Just as free programs need free libraries, free machines need free electronics and ultimately free components