Like other nonexcludable goods, content

lends itself to FREE RIDING.

Here, we compare these contexts using the

GARN prism, identify the threats associated ^{Nonexcludable High} with the GARN phe- ^Low nomenon, and discuss ways to counter the threats. We provide a unified framework to phenomena that were previously viewed in isolation. Our aim is to unravel the common thread among Internet-facilitated contexts and consequently examine how lessons from one context may be used in others. Before exploring specific GARN phenomena, we turn to the economics literature and present several types of goods, as they are useful in our analysis.

Goods can be classified into four categories across the dimensions of nonexcludability and nonrivalry (see Table 1) [ 3]. For example, it is difficult to prevent nonexcludable goods (such as public parks) from being consumed by nonpayers; a nonrival (or nonconsumable) good (such as a TV show) can be consumed without diminishing its availability for future consumption. Unlike physical goods, information goods (such as software and content) are typically nonrival [ 11].

OSS development is a growing phenomenon [ 12], evident in the number of users registered at Source-forge.net (a host for open-source projects, www.sourceforge.net) that increased from around 500,000 in 2003 [ 6] to more than 1. 8 million in 2008. A growing variety of software applications are developed in the open-source model; more than 170,000 are registered at Sourceforge. Contributions are made by individuals independently deciding to donate time and effort to produce software that is freely available to all (a pure GARN) or by software companies actively supporting OSS development.

The availability of high-speed networks allows the sharing of computer resources via grid computing,

Table 1. Types of goods [ 3].

enabling better allocation

Nonrival

and use of computing ^High resources within and ^{Public good} among organizations. ^{Collective good} Grids may be closed (to members of a certain community) or open (accessible to all); they may also be one-way (contributors lack access to the pooled resource) or two-way (contributors are also receivers). A well-known example of a grid-computing project is SETI@home, a volunteer-based grid comparable to the strongest computer available today, for searching for “extraterrestrial intelligence.” Computing grids can be viewed as grids of computers and primarily represent an architectural issue. However, our focus here is the wider issue of grids of computer users and owners, including their social and organizational implications.

Increasing use of wireless broadband computing is accompanied by the emergence of wireless commons, whereby participants provide Internet access to other participants through their WLAN access points [ 2]. Like computing grids, wireless commons may be either closed or open to all [ 2]. If the latter, which represents a pure GARN mechanism, individuals must not password-protect their access points.

Content GARN often manifests itself as contribution of content produced by others (such as media files shared via P2P systems). However, in many P2P networks, contributing and consuming are technically almost inseparable for users. Therefore, we won’t analyze P2P networks here further. A clearer example of content GARN is knowledge-based, self-created content, such as Wikipedia, a free online encyclopedia with 1. 4 million entries in English alone, written and edited by volunteers [ 7].

The GARN mechanism carries with it a number of threats that can be categorized using the goods typology (see Table 2); these are analyzed in the next sections. Given the common GARN mechanism

Low

Commons good