1. Introduction

A central problem of computer science is the integration of knowledge and coordination of action in complex systems. The same may be said of society. In society, however, this problem has been faced for millennia rather than decades, and diverse solutions have been tested for effectiveness through hundreds of generations of competition. Efforts to understand the resulting institutions and to describe their principles of operation have spawned the science of economics.

Contrary to common impressions (fostered by media coverage of politics and the stock market), most economic inquiry has little to do with guessing economic trends. Economics has many branches; the branch most relevant to this paper studies the consequences of pursuing goals within the constraints of limited knowledge and resources, and studies the institutions and patterns of behavior adapted to this pursuit. This branch of economics can without embarrassment be termed a science, since it meets the criteria for a scientific discipline [1,2].

At the broadest level of abstraction, the problems of social and computational coordination are fundamentally similar. Concrete parallels, however, are rough: memory space is a bit like land, or perhaps a raw material; processor time is somewhat like labor, or like fuel; software objects are like workers, or perhaps like managers or firms. In [I] we list a number of fundamental differences between computational and human markets. For example, within a computational system, activities need produce neither pollution nor other effects on non-consenting objects; the most typical product, information, does not form a depletable physical inventory; specialized labor forces (copies of specialized objects) can be expanded almost instantly and can be cut back without human anguish.

Despite these deep differences, we argue that the fundamental parallels between the problems of social and computational organization are strong enough to motivate the wholesale importation of economic models and metaphors into the computational domain, at least on a trial basis. These differences do, however, suggest that forms of organization that fail or are rejected in one domain may prove workable and desirable in the other. For example, the ability of computational systems to establish rules as genuine constraints where an analogous human legal system can only penalize violations makes possible patterns of organization that can only be approximated in society.

1.1. Why focus on markets?

For a variety of reasons, this work explores essentially pure markets as models of economic organization for computation, supported by a minimal "legal" framework of foundational constraints. A large body of economic theory and historical experience indicates that markets are, on the whole, remarkably effective in promoting efficient, cooperative interactions among entities with diverse knowledge, skills, and goals. Historically, those entities have been human beings, but economic principles extend to decision-making agents in general and hence to software objects as well. In [I], markets are considered as ecosystems and compared to others, such as biological ecosystems. This examination shows how the distinctive rules of markets (such as the suppression of force and protection of trademarks) foster the spread of cooperation (and encourage entities to compete to be effective cooperators).

This paper argues that market ecosystems are particularly appropriate as foundations for open systems [II], in which evolving software spread across a distributed computer system serves different owners pursuing different goals. When also open to human society, computational market ecosystems will enable diverse authors to create software entities and receive royalties for the services they provide, and enable diverse users to mold the system to their needs by exercising their market power as consumers. Computational markets can be made continuous with the market ecosystem of human society.

1.2. Sketch of a computational market

This line of investigation leads us to propose what may be called the agoric approach to software systems. Agoric stems from agora, the Greek term for a meeting and market place. An agoric system is defined as a software system using market mechanisms, based on foundations that provide for the encapsulation and communication of information, access, and resources among objects. Each of these notions plays a role in supporting computational markets.

Here, the notion of "object" is independent of scale and language, and includes no notion of inheritance. An object might be small and written in an object-oriented language; it might equally well be a large, running process (such as an expert system or a database) coded internally in any manner whatsoever. Objects are assumed to communicate through message passing and to interact according to the rules of actor semantics [3,4], which can be enforced at either the language or operating system level. These rules formalize the notion of distinct, asynchronous, interacting entities, and hence are appropriate for describing participants in computational markets.

Encapsulation of information ensures that one object cannot directly read or tamper with the contents of another; communication enables objects to exchange information by mutual consent. The encapsulation and communication of access ensures that communication rights are similarly controlled and transferable only by mutual consent. These properties correspond to elements of traditional object-oriented programming practice; in large systems, they facilitate local reasoning about competence issues-about what computations the system can perform. Extending encapsulation to include computational resources means holding each object accountable for the cost of its activity; providing for the communication of resources enables objects to buy and sell them. In large systems, these extensions facilitate local reasoning about performance issues-about the time and resources consumed in performing a given computation. Computational foundations suitable for markets thus offer advantages in the performance domain like those offered in the competence domain by object-oriented programming.

For concreteness, let us briefly consider one possible form of market-based system. In this system, machine resources-storage space, processor time, and so forth-have owners, and the owners charge other objects for use of these resources. Objects, in turn, pass these costs on to the objects they serve, or to an object representing the external user; they may add royalty charges, and thus earn a profit. The ultimate user thus pays for all the costs directly or indirectly incurred. If the ultimate user also owns the machine resources (and any objects charging royalties), then currency simply circulates inside the system, incurring computational overhead and (one hopes) providing information that helps coordinate computational activities.