Protecting Open Innovation: The Defensive Patent License as a New Approach to Patent Threats, Transaction Costs, and Tactical Disarmament

Historically, Open Innovation Communities (OICs), such as those who make free and open source software (FOSS), have opted out of the patent system for three main reasons: Patents are expensive to acquire and enforce; they are philosophically, culturally, and politically anathema to many OIC members; and even when they appear to be acquired for “defensive�? or other altruistic purposes, there has been no guarantee against someone later “weaponizing�? them for use in an offensive attack. In this article, we first analyze how and why various OICs have approached strategic concerns about intellectual property and then make a case for why OICs should – and indeed must – opt back into the patent system if they wish to protect themselves from the growing threats that patents pose, including threats from the increasingly complex “thicket�? of patents on software, standards, and ecosystems that have become the battlegrounds over technologies such as smartphones, online video, and social media. We support our argument with a proposal for a new OIC patent license – the Defensive Patent License. Built on legal and normative approaches similar to the GNU General Public License and Creative Commons licensing, the DPL directly addresses the three reasons for OIC opt-out and provides several positive reasons for opting in, including increasing access and interoperability between OIC technologies, improving the prior art, imposing legally-binding restrictions that require DPL patents to remain “defensive,�? and preventing patent trolls, “bullies,�? or others from ever asserting DPL patents against those who use the license. We end with a comparison of the DPL to other OIC patent strategies to situate it within the current set of options OICs face in dealing with patent threats.


I. INTRODUCTION
Historically, Open Innovation Communities ("OICs"), 1 ranging from those who make free and open source software ("FOSS") projects to various do-it-yourself hardware manufacturing, 2 3D printing, 3 synthetic biology projects, 4 and green technologies, 5 have struggled to 1. We define OICs as organizations and collections of individuals that value sharing technologies and information as broadly as possible with a substantial commitment to user freedoms. As discussed below, OICs adopt this strategy for a variety of economic and cultural reasons. These companies, academic institutions, nonprofits, and project-based communities generally either grew out of the Free and Open Source Software ("FOSS") movement or for strategic or mission-based reasons adopted FOSS-inspired  tions to patents that cover software, business methods, and the like as ethically inappropriate or scientifically questionable attempts to own and control mathematical formulas and other laws of nature. 12 Third, even when patents appear to be acquired for "defensive" or other altruistic purposes, there has been no guarantee against someone "weaponizing" them in the future for use in an offensive attack. 13 A range of companies, especially those in the software, hardware, and semiconductor fields, have described their patent portfolios as "defensive," and in some cases have made promises to this effect; however, these companies have not always made legally binding commitments to keep those promises. 14 Some of these companies have then allowed "defensive" patents to be used or sold for offensive purposes or have been forced by investors or through bankruptcy proceedings to sell them to entities that might use them for offensive purposes. 15 Even socalled "patent pledges" not to assert patents offensively have raised significant questions as to their enforceability and scope. 16 This has led to further trepidation about patenting for OICs, especially with the rise of relatively new patent portfolio owners whose intentions may not align with the previous patent portfolio owners. 17 12. See BEN KLEMENS, MATH YOU CAN'T USE: PATENTS, COPYRIGHT, AND SOFTWARE 44-72 (2005), available at http://www.e-reading.org.ua/bookreader.php/133502/Math_ You_Can%60t_Use%3A_Patents,_Copyright,_and_Software.pdf; Richard Stallman, Patent Absurdity, GUARDIAN (June 22, 2005) Defensive" Patent Has Chilling Effect, H OPEN (Mar. 16, 2009, 1:37 PM), http://www.h-online.com/open/news/item/Red-Hat-defensive-patent-has-chilling-effect-740575.html. Another company, RPX, buys up or licenses patents and then licenses them out to companies that sign up for a substantial fee. RPX describes itself as a "defensive patent aggregator," though it has been criticized as little more than a "troll" that charges companies for protection against suit.  TECHCRUNCH (Nov. 24, 2008), http://techcrunch.com/2008/11/24/is-rpxsdefensive-patent-aggregation-simply-patent-extortion-by-another-name.
15. See, e.g., Alorie Gilbert, Commerce One Patent Sale Has Google Worried, ZDNET UK (Nov. 25, 2004)  In this Article, we seek to question and reinvigorate the debate over if, when, how, and why OICs might use patents. We start by analyzing the three main reasons why OICs opt out of the patent system. We then look at how various OICs have approached strategic concerns about intellectual property and derive a framework for evaluating OIC patent defense strategies.
We then make a case for why OICs should -and indeed mustopt back into the patent system if they wish to protect themselves from the growing threats that patents pose, including threats from the increasingly complex "thicket" of patents on software, standards, and ecosystems that have become the battlegrounds for technologies such as smartphones, online media, and social media. 18 To be clear, we are certainly sympathetic to OICs' concerns about patent subject matter and scope; effectively addressing the breadth of patents could address many concerns about patent threats. But the Supreme Court's decisions in Bilski v. Kappos 19 and Mayo v. Prometheus 20 indicate that software and some other abstract patents are here to stay for the foreseeable future. Similarly, the America Invents Act 21 shows that Congress is also unlikely to fundamentally reform patentable subject matter or patent scope anytime soon. Given this, making the system as it exists safer for innovation may be the most practical approach for now.
We support our argument with a proposal for a new OIC patent license -the Defensive Patent License ("DPL"). The DPL is a form of patent license that serves as the connection point for a distributed defensive cross-license network. Built on legal and normative approaches similar to the GNU General Public License 22 and Creative tel Patent Auction Goes to Apple/Microsoft/RIM Consortium, UNOFFICIAL APPLE WEBLOG (July 1, 2011, 6:45 AM), http://www.tuaw.com/2011/07/01/nortel-patent-auction-goes-toapple-microsoft-rim-consortium; When Patents Attack!, THIS AMERICAN LIFE (July 22, 2011), http://www.thisamericanlife.org/radio-archives/episode/441/when-patents-attack.
18. See, e.g., Ingrid Lunden, Spotify Gets Hit with a Patent Suit from Nonend, a Dutch Peer-to-Peer IP Holder, TECHCRUNCH (Aug. 15, 2012), http://techcrunch.com/ 2012/08/15/spotify-gets-hit-with-a-patent-suit-from-nonend-a-dutch-peer-to-peer-ip-holder; Baio, supra note 13; Press Release, Public Knowledge [Vol. 26 Commons licensing, 23 the DPL directly addresses the three reasons for OIC opt-out and provides several incentives for opting in, including (1) increasing access and interoperability between OIC technologies, (2) improving the prior art, (3) imposing legally binding restrictions that require DPL patents to remain "defensive," and (4) preventing patent trolls, "bullies," 24 or others from ever asserting DPL patents against those who use the license. 25 We end with a comparison of the DPL to other OIC patent strategies.

II. OPEN INNOVATION PATENTING: BENEFITS AND BARRIERS
To better understand why OICs opt out of the current patent system, we first examine one of the primary strategic reasons that proprietary innovation communities patent: defensive patenting. Below we examine the costs and benefits of this strategy and consider additional reasons why OIC firms have largely failed to adopt a similar approach.

A. The Costs and Benefits of Defensive Patenting
In its most basic form, defensive patenting is the practice of seeking patents in order to deter offensive lawsuits rather than to gain more traditional patent benefits, such as seeking licensing revenue, facilitating joint research and development, or excluding competitors. 26 To engage in defensive patenting, competitors within an industry each build significant patent portfolios and then threaten to use those portfolios in response to offensive patent threats. 27 The greater the size and scope of a given firm's portfolio, the more dangerous and costly it is to litigate against, thus serving as an ex ante deterrent. 28 These costs include not only those of litigating the patents, but also 23. See About, CREATIVE COMMONS, http://creativecommons.org/about (last visited Dec. 22, 2012).
25. We recognize, of course, that our proposal may be the second-best solution and that direct reform to the patent system through the courts, Congress, or the U.S. Patent and Trademark Office might be preferable. However, unless and until such reform occurs, we believe private ordering solutions such as the DPL can continue to improve the patent situation for OICs and provide concrete social welfare benefits.

No. 1]
Protecting Open Innovation 7 the high information costs associated with evaluating what are often dozens if not hundreds of opposing claims on both infringement and invalidity grounds. 29 In well-resourced and patent-heavy industries, such as semiconductor manufacturing, 30 these costs and the potential downsides of litigation 31 can lead to cross-licensing détentes, with significant players either preemptively agreeing to avoid litigation or leveraging defensive patents to resolve disputes after litigation has commenced. 32 In light of these benefits, the practice of building a defensive patent portfolio is a common intellectual property strategy of many proprietary innovation firms. It is worth noting, however, that defensive patenting does not work against all potential patent plaintiffs. While it has generally been effective against competitors who share the same high costs of litigation, such as so-called patent "bullies" who might threaten smaller competitors or new entrants to a field, 33 it has generally failed to deter (and in some ways may have even inspired) so-called "patent trolls," who often acquire otherwise inactive patents and file or threaten lawsuits for the sole purpose of generating licensing income. 34 Because trolls cannot be countersued -they do not make, use, import, sell, or  Patent Troll Myths, 42 SETON HALL L. REV. 457, 462-66 (2012), we use the term generally to mean patent litigation plaintiffs that have neither practiced, researched, nor developed the technology claimed in the patent-in-suit in any significant or substantial way. While others prefer the term "non-practicing entities," id., we find that term too limited in some contexts, such as academic research institutions. offer for sale anything that could be infringing, inoculating them against suit -a defensive portfolio has little effect against them. However, the patents in a defensive portfolio, if later acquired by a troll, could become weapons to wield against other defendants. The defensive patenting strategy works for traditional proprietary innovation firms, in part, because the most significant costs -patent prosecution, portfolio management, litigation, and licensing negotiation -can be concentrated relatively easily within a firm's legal department. Microsoft can justify the cost of filing for 3000 patents per year because the benefit of having 3000 potential counterstrikes against a future patent threat presumably outweighs those costs. 35 Microsoft can also internalize both the costs and benefits. For example, Microsoft can afford the patent application and maintenance fees, the cost of tracking and maintaining its portfolio, the cost of patent prosecution and licensing advice, and it can hire in-house patent counsel to coordinate patent strategy efficiently and under the protections of the attorney-client privilege, all of which suggest economies-of-scale efficiencies. 36 The benefits are also clearly definable and to some degree measurable -a reduction in the costs imposed on the firm by patent litigation.
OIC firms, however, have much more difficulty achieving this balance. First, individual OICs often lack the legal and fiscal resources to build and enforce large patent portfolios, to absorb the information costs of understanding oppositional portfolios, or to incur the transaction costs of forming cross-license agreements. 37 While some larger OIC players -Google, Novell, IBM, and Red Hat in particular 38 -have both in-house and outside patent counsel to shoulder some of this burden, many of their collaborators within the Linux operating system, open web, and Android mobile OS ecosystems are either volunteer projects, nonprofit foundations, or startups, and cannot absorb these costs. Second, many OICs, especially FOSS projects, are decentralized or organized outside of formal corporate structures, and thus incur significant coordination costs whenever there are efforts to collaborate across projects or firms. This is particularly true when it comes to legal strategy, as such projects tend to 35 have few, if any, lawyers on staff. This is one reason why OICs rely on distributed and decentralized means of production to create technologies. By employing distributed production techniques that operate via purposeful openness and sharing, they are able to distribute the costs and benefits throughout the community and take advantage of the network effects of their production processes.
In the legal arena, OICs have also sought to decentralize costs and to use network effects to maximize benefits by employing scalable and standardized "off-the-shelf" intellectual property licenses that do not require downstream lawyer involvement. This is in part because decentralized production strategies present some significant challenges. For example, obtaining legal advice under the attorney-client privilege is far more difficult when the engineers who will implement the advice are volunteers who may be distributed over dozens of physical locations worldwide and who mainly communicate through public email lists or centralized code repositories. 39 This is one reason why the use of standardized licenses such as the GPL and Creative Commons have been so popular among OICs. Once developed, they provide scalable legal tools to promote decentralized participation and administration of creative production without requiring significant additional legal or fiscal resources. 40 However, these strategies have some significant limits when case-by-case legal questions or transactions arise.
Traditional defensive patenting presents such a situation, as it does not scale as easily as copyright licensing. Unlike obtaining a copyright, obtaining a patent is not automatic, and involves significant prosecution and administration costs. 41 On top of this, the costs of patent portfolio management, analysis of oppositional claims and likely threats, licensing and negotiation, and litigation are all much more significant for patents than for copyrights. 42 39. See, e.g., phpWebSite Content Management System Forums, SOURCEFORGE, http://sourceforge.net/projects/phpwebsite/forums (last visited Dec. 22, 2012) Noga, 168 F.3d 1282, 1288(11th Cir. 1999). 42. The very nature of patent infringement differs from copyright infringement in ways that make decentralized legal defense of patent threats much more challenging. With copyright, one cannot be liable for infringement without either evidence of literal copying or access to the plaintiff's work coupled with evidence of substantial similarity between the two works. E.g., Three Boys Music Corp. v. Bolton, 212 F.3d 477, 481 (9th Cir. 2000). This means that OICs that develop original computer code are quite unlikely to be accused of copyright infringement. By contrast, one can infringe patents unknowingly, even when writing original code. Global-Tech Appliances, Inc. v. SEB S.A., 131 S. Ct. 2060, 2065n.2 (2011 ("[A] direct infringer's knowledge or intent is irrelevant."). This makes it much more difficult for OICs to predict and avoid patent infringement claims. As we discuss below, the [Vol. 26 Thus, in order for OICs to take advantage of a defensive patent strategy, that strategy would have to address these costs and concerns and ensure that enough benefits accrue to the OIC to outweigh them. In particular, given the general distributed structure for OIC costs and benefits, a successful defensive patent strategy must either take advantage of the same distributive economics or somehow compensate for the lack of traditional firm concentrations in these areas.

B. The Cultural and Political Barriers to OIC Patenting
Cost-benefit challenges are the most straightforward reason why OICs might have shown limited interest in opting in to the patent system, but there are additional significant reasons, including the strong cultural and political preferences against patents held by many OIC participants. OICs have historically disfavored patents, especially on software and information goods. 43 As described further in Part II, for both philosophical and practical reasons, many in OIC communities reject the exclusive, proprietary models of innovation and production that patents have traditionally undergirded. It is common for OIC advocates to consider patents anti-competitive, anti-innovative, and antithetical to rapid development and to the spirit of openness and freedom. 44 While this is certainly true of a strong contingent in the FOSS community, it is also true within many parts of the proprietary software community. 45 TECHDIRT (Aug. 23, 2011, 9:46 AM), http://www.techdirt.com/articles/20110822/13094215621/what-idiot-wrote-patent-thatmight-invalidate-software-patents-oh-wait-that-was-me.shtml.

No. 1]
Protecting Open Innovation 11 dio ("NPR") report, provides an example of the strong perception that patents are a negative force in the software and information industries: [W]e talked to a half dozen different software engineers. All of them hated the patent system, and half of them had patents in their names that they felt shouldn't have been granted. In polls, as many as 80 percent of software engineers say the patent system actually hinders innovation. It doesn't encourage them to come up with new ideas and create new products. It actually gets in their way.
Many patents are so broad, engineers say, that everyone's guilty of infringement. This causes huge problems for almost anyone trying to start or grow a business on the Internet. 46 As suggested by the NPR story, longstanding questions about the status of software and other informational innovations as patentable subject matter, and uncertainty concerning the boundaries of patent protection in these areas, have likely contributed to these negative perceptions of patents in OICs. In general, laws of nature, discovered facts, and abstractions (such as mathematical equations) are unpatentable, owing both to the constitutional origins of patenting and to concerns over the impact that patents on broad principles and ideas could have on downstream innovation. 47 How software fits into patentable subject matter has been contested since its entry into the modern consumer marketplace, and new products and services periodically reinvigorate the longstanding debate among the federal judiciary, intellectual property policymakers, and academic commentators about the extent information products and services should be patentable. 48

46.
When Typically, OIC firms have come down strongly on the side of limiting patentable subject matter in this area, in part because they have not traditionally relied upon patents as incentives for their own production, and in part because they fear software patents being wielded as a mechanism for reducing their ability to innovate freely.
OIC cultural objections to software and other abstract patents developed in parallel with case law that established software as patentable subject matter. In a series of highly contested rulings through the 1970s and 1980s, the Supreme Court struggled with the subject matter issue, 49 eventually deciding 5-4 in the 1981 case Diamond v. Diehr that software that "transform [ed] or reduc[ed] an article to a different state or thing" could be patentable. 50 For over a decade after Diehr, the patent-specialist Federal Circuit Court of Appeals engaged in its own internal battle over the limits of software patentability. It eventually decided en banc in the 1994 case In re Alappat that a generalpurpose computer running novel and nonobvious software was a patentable "new machine" under the statutory language of the Patent Act. 51 After Alappat, the Federal Circuit allowed patents on increasingly more abstract inventions, culminating in State Street Bank & Trust Co. v. Signature Financial Group, a 1998 case in which the court held that a hub-and-spoke method for mutual fund investing was patentable. 52 This resulted in the U.S. Patent and Trademark Office ("USPTO") granting a set of broad, criticized patents -for example, a patent to Amazon on one-click online shopping 53 -resulting in a backlash in the information industries 54 and at least one Supreme Court dissent that appeared to question the Federal Circuit's approach. 55 In 2008, the Federal Circuit stepped back somewhat from State Street, holding en banc in In re Bilski that a patent on a method of hedging financial risk in commodity trading that claimed neither a ert Merges, while remaining agnostic to software and other abstract patents on subject matter grounds, argues for reforms in the patenting process in order to weed out overbroad or otherwise poor-quality patents. See Merges, supra note 11, at 598. And others argue strongly for the patentability of software.

No. 1]
Protecting Open Innovation 13 "new machine" nor a transformation of matter, was a bridge too far. 56 This decision fueled hope in OICs that patents on software or other abstract information technologies and services might finally be disallowed, or at least narrowly granted. 57 On appeal, however, the Supreme Court shifted away from this approach and found the Bilski patent invalid under a different rationale: that it was too abstract, and thus non-patentable subject matter. 58 This left the patentability of software in much less doubt. Therefore, while skepticism in OICs about the appropriateness of software and information patents remains, 59 they nonetheless must face the threat presented by such patents.
There are also strong objections within OICs to software and information patents' perceived lack of quality and over-reaching breadth. In particular, many OIC advocates have expressed concern that the USPTO issues patents in this sector on incremental improvements that should be considered obvious and grants overbroad claims that allow patentees to threaten entire industries. 60 This perception that software patents are illegitimate as legal instruments continues today, erupting every few weeks or months when the next objectionable patent issues from the USPTO. 61 Moreover, the rise of patent trolls has further increased hostility to the idea of patents as beneficial legal tools. 62 As stories of patent threats against information innovators continue to be reported in the press -especially stories of threats based on patents of perceived lower quality -the attitude toward software patents has continued to sour among developers who already believed that companies should compete on other grounds, such as branding, customer support, quality, and first-mover advantages. 63  strategy to be effective, it would have to address these concerns as well.

C. The Reliability of OIC Patenting Promises
Beyond cost concerns and cultural objections, OIC firms are also concerned about the reliability of any promise made by patent holders to use patents for altruistic purposes, such as defensive use only. Patents used only for defensive purposes -in other words, patents that should not pose any threat to OIC firms that do not assert patents themselves -are in theory far less objectionable to the OIC community. Still, concerns remain within OICs that patents allegedly -or initially -prosecuted for defensive purposes will eventually be used offensively. The original patent holder might have a change of heart or business strategy; even if this does not happen, some later successor may choose not to honor a predecessor's pledge. 64 Where a patent is acquired through a hostile takeover or a bankruptcy proceeding, the original pledge may be particularly vulnerable. For example, as part of the bankruptcy proceeding for web services firm Commerce One, thirty-nine patents that covered basic web protocols went on the auction block, causing concern in Silicon Valley about who would acquire them. 65 The patents were sufficiently broad to generate fears that a patent troll or a firm with anti-competitive intentions would prevail in the auction and wreak havoc on various aspects of e-commerce. 66 In addition to entering bankruptcy, firms regularly sell assetsincluding patents -merge with other entities, or are acquired. 67 Since its inception over a decade ago, Intellectual Ventures, a firm focused on generating revenues from patent licensing rather than making and selling products, has purchased over 30,000 existing patents. 68 Nortel Networks recently auctioned off over 6000 patents on its way to shut-

No. 1]
Protecting Open Innovation 15 ting down. 69 These cautionary tales continue to raise questions about how permanent any particular commitment to defensive patenting can be. 70 Unless OICs can depend on patents remaining defensive, it is unlikely they will invest in a defensive patent strategy. From the above analysis, we see at least three substantial barriers to defensive patenting as a strategy for OICs: (1) overly concentrated costs and benefits from patenting, (2) a perceived mismatch between OIC normative values and patenting (and in particular, the patenting of information goods such as software), and (3) a perceived lack of reliability of patent owners to honor existing commitments to use patents defensively. We believe that any defensive patent strategy that will be effective for OICs must address each of these barriers. In the next Part, we discuss these strategic criteria in relation to existing OIC theory and practices, and use the resulting analysis to develop a more detailed framework for effective OIC patent defense strategies.

III. A FRAMEWORK FOR EVALUATING OIC PATENT STRATEGIES
Exploring the barriers to defensive patenting, then, helps explain its limited present benefit to OICs. Turning these barriers around, we can identify the following criteria for a viable OIC patent strategy: (1) distributed costs and benefits where the collective benefits outweigh the collective costs, including fiscal, legal, information, and transaction costs; (2) cultural and political alignment with OIC values; and (3) reliable commitments to these values and, importantly, to the use of patents solely for defense. In this Part, we expand upon these criteria in two directions to develop a more detailed framework for discerning what an effective OIC patent defense strategy should look like. First, we examine the broader OIC literature on intellectual property strategies generally, and FOSS literature and licensing practices more specifically. Second, we examine existing OIC patent defense strategies for common values and techniques that have proven acceptable and effective. The theories, values, and practices of OICs are described in an increasingly robust set of academic literature and practical resources. In this Section, we draw from these resources to identify framework components that address each of the three barriers to defensive patenting we have identified. Two of these barriers -cost-benefit concentration and a lack of reliably defensive patents -each have direct analogues in the FOSS communities' practical methods for addressing copyright issues and organizing production. Addressing the third -a mismatch between patenting and OIC community values and normative theories -requires inquiry into those values and how OIC communities have addressed previous concerns over their mismatch with copyrighting software, eventually adapting and embedding them into production practices.
We begin with OIC community values and organizational theory. In the United States, the dominant theory of intellectual property is based on providing economic incentives through the artificial creation of exclusivity. 71 Informational goods are by nature non-exclusive, meaning that they generally lack barriers to sharing, 72 and nonrivalrous, meaning that one person's use does not reduce the enjoyment of the informational good by other users. 73 Under the dominant theory, legally created exclusivity -in the form of copyrights and patents -creates rivalry and gives competitors an incentive to innovate because they can expect to exploit the exclusivity to recoup their investment, and because others' exclusive rights dictate that they cannot simply copy prior goods without permission and, often, payment. 74 Without this exclusivity, so the theory goes, firms will be unable to recoup the costs of development and reap profits in the marketplace; therefore, they will lack motivation to innovate. 75 OICs have long challenged the dominant incentive theory. While individuals, groups, and companies within the different OIC communities have varying viewpoints, all directly reject the need for infor- Other commentators describe a variety of production models based on motivations other than IP-based exclusivity. 83 Academic theorist Yochai Benkler points out that the dominant exclusivity theory is drawn from the industrial model of production prevalent in the nineteenth and twentieth centuries, and argues that "non-proprietary strategies have always been more important in information production than they were in the production of steel or automobiles." 84 Benkler agrees with FOSS commentators that motivations go beyond economic incentives based on exclusivity, and argues that other motivations -ranging from benefits to reputation, to the importance of developing know-how, to stimulating demand for a related product (such as attorney services), and so on -represent a large fraction of the motivations behind information production. 85 For members of OICs, intellectual property's artificial exclusivity is thus often seen as a barrier rather than a benefit. Copyrights, which make the rights to reproduce, create derivative works, and distribute software exclusive to the copyright holder, and software patents, which make the rights to make, use, sell, import, or offer for sale the software exclusive to the patentee, 86 severely undermine the expectations and values described by FOSS commentators and embraced by OICs more generally. The programmer who wants to "scratch an itch" is stymied by a prohibition on copying software that might need modification or that provides a needed component for addressing the "itch." With patents, this prohibition is even stricter, prohibiting even independent implementations of particular software methods. For the free software programmer who wants to "spread[] freedom and cooperation," 87 the exclusive rights of intellectual property stymie the ability to develop and share solutions and efficiently cooperate across decentralized networks. 88 This mismatch between traditional intellectual property theory and the goals of FOSS developers contributes to the most basic norms of the OIC communities: the norms of openness and freedom. OIC leaders often pronounce these norms 89 in such statements as the FSF's Free Software Definition ("FSD") 90 and the Open Source Initiative's Open Source Definition ("OSD"), 91 both of which require distribution of source code (openness) 92 and the freedom to use, modify, and redistribute software for any purpose (freedom). 93  The norms of freedom and openness find their practical expression in the terms of the standardized software licenses that meet the FSD's or OSD's requirements. The GNU General Public License, Creative Commons licenses for creative works, and the many software licenses that comply with the OSD are examples. 95 Though they vary considerably in the details, these licenses all grant the rights to use, modify, and redistribute the software they cover (thus supporting the norm of freedom), and require that the human-readable source code be distributed along with the machine-readable compiled code (thus supporting the norm of openness). Therefore, these licenses use the licensor's copyright rights -and in some cases, contract law 96to allow and enforce openness and freedom. Further, the licenses themselves are all transparent and publicly available -anyone can review the terms, at any time -further supporting the norm of openness. 97 The method of using standardized copyright licenses to encourage and enforce OIC values and norms also operates to lower and decentralize the costs of production and to distribute the benefits across the entire OIC community. At one level, the standardized nature of the licenses reduces the fiscal and legal resources required to develop and apply the license -writing a standardized license limits these costs to the initial drafting and periodic updates, as everyone in the community simply uses the form license when distributing code. 98 This method also drastically lowers the informational and transactional costs of licensing -the terms are standard and transparent to all parties, so there is no information asymmetry and no need to negotiate terms. This reinforces openness at both the legal layer and the code layer. At the next level, open licenses allow for the decentralized production of software and other informational goods. Because the intellectual property rights are standard, and freedom to modify is included in these open licenses, anyone can contribute to a project without worrying about licensing rights -lowering production costs and generating positive network effects that can enhance production speed and quali- ty. 99 Finally, the right to use or modify the software is available to anyone, fully distributing the benefits as part of the norm of freedom.
Finally, FOSS licenses provide reliability to OIC communities. Because the community's norms are reflected in the licenses' requirements, they are seen as legally enforceable, 100 and thereby reliable. A FOSS developer knows that the source code needed to understand and modify a product must be included with the compiled code because the license requires it. Communities that support "copyleft" principles go further, and use licenses to create reliable sharing requirements. A copyleft license not only allows modification and requires that the source code be available, it also requires that any modifications be distributed under the same license, making the freedom and openness requirements "viral." 101 The GNU General Public License and the Creative Commons Share-Alike licenses are the canonical examples. By conditioning the license terms upon the licensee's commitment to freedom and openness, copyleft OICs can ensure that, as the network of users grows, new users enhance the value of the network through ongoing commitments to the same principles and norms. This fosters reliability by creating enforceable norms and limits gamesmanship. Creating a legal requirement to share, via the copyright license, is the mechanism the copyleft FOSS communities use to ensure that a software project, and any distribution based on it, remains reliably open and free. As the FSF states: "If middlemen could strip off the freedom, we might have many users, but those users would not have freedom . The theory and practice behind OICs, especially FOSS communities, give us several principles and methodologies to draw upon in considering how to overcome barriers to designing an effective framework for OIC patent defense.

A. Principle One: Commitment to Freedom i. Freedom to Operate and Freedom to Innovate
The first and foremost normative principle in OICs is freedom, enforced by cultural expectations and licensing regimes. Specifically, most OIC norms support two types of freedom: freedom to operate and freedom to innovate. Freedom to operate, a term most often used in the patent world, means a general freedom from intellectual property limitations to make, use, sell, import, or offer for sale a particular device or service. 103 For most FOSS copyright licenses, this concept translates to the freedom to run the program for any purpose, and the freedom to study how the program works. The freedom to innovate extends this to include the freedom to change the software to make it do what you wish, 104 including the freedom to modify and redistribute the software. 105 In the realm of patents, we believe the same principles should be supported. The freedoms to operate and innovate benefit OICs by encouraging collaboration, lowering barriers to entry, and generally making technological innovations more accessible. Interestingly and importantly, the right to modify in copyright is specifically granted to copyright owners via the derivative work right; 106 however, in patent law, studying inventions and innovating on top of them is generally not an exclusive right. 107 Thus, in some ways, patent law is more initially amenable to this second freedom than copyright. A second principle of OIC culture is openness. Openness primarily refers to access to sufficient knowledge about an information good to understand it and to share that information with others as desired; however, it can also mean transparency of the legal regime governing the good.

i. Access to Knowledge
The first component of openness is meaningful access to knowledge about the information good. The primary means by which OICs commit to this component of openness is by sharing the basic building blocks of the technology and then enforcing that sharing via open licenses. For example, in software, the shared building blocks primarily consist of the source code; in hardware, the schematics that describe the hardware; in biology, the genetic sequences that make up an innovative compound; and so on. 108 For OIC patent strategies, access to knowledge is an essential normative requirement. Permission to use patented technology without full information about the technology's characteristics would undermine this principle. In fact, patent law purports to serve this exact purpose -disclosure of knowledge about the invention -as the quid pro quo for exclusivity. 109 However, some commentators argue that under current patent law, this goal has been undermined by several doctrines; chief among these are the danger of additional willful infringement damages that may stem from reading patents that may cover one's product 110 and the generally weak enforcement of the written description and enablement requirements, 111 which are intended to serve as mechanisms for ensuring robust disclosures. 112 That said, were innovators encouraged to read patents rather than avoid them, and were the written description and enablement requirements

ii. Transparent and Non-Discriminatory Legal Terms and Enforcement Conditions
The second aspect of OIC openness moves beyond the technology itself to the transparency and non-discriminatory nature of the legal terms and enforcement conditions of standardized OIC licenses. The openness of OIC licenses allows every member of an OIC to read, understand, and evaluate the legal instruments used to enforce OIC norms. Broad publication and replication of OIC licenses removes the need for individualized license negotiations, thus encouraging the use of OIC technologies and participation in the community by lowering individual information and transactional costs. 113 Open licensing also facilitates trust and accountability among OICs by providing a common set of legal commitments that can be tested both in court and in public discourse. Additionally, open licensing can reinforce various non-monetary community norms, such as sharing and collaboration, by requiring the general inclusion of authorship information. 114 Thus, any OIC patent defense strategy should also provide similar legal transparency and non-discrimination.

iii. Interoperability and Technology Neutrality
The third and fourth aspects of openness are interoperability and technology neutrality. Because these two aspects of openness go hand-in-hand, we have grouped them together. Interoperability is a design principle that encourages technologies to allow interaction with other technologies. 115 In this sense, "open" means both transparent and accessible to others who might want to interact and build upon the technology. When successful, interoperability can create network effects that increase the value of any given OIC input to every other aspect of the system. For example, allowing a technology to work with many different operating systems, keyboards, monitors, or mobile devices ensures that the technology has the greatest potential for

Harvard Journal of Law & Technology
[Vol. 26 increasing social welfare, and allows for robust competition among a range of aftermarket options. 116 Technology neutrality is a related concept that is generally understood as a principle of non-discrimination. 117 The OSD addresses technology neutrality with a simple rule: "No provision of the license may be predicated on any individual technology or style of interface." 118 In other words, the OSD prohibits an innovator from granting license rights only to those who use some forms of technology and not others; rather, the license rights must apply to all technologies. This ensures that OIC technologies cannot foreclose competing or new technologies, thus maximizing access and enhancing network effects.

C. Principle Three: Distributed Costs and Benefits
A third principle focuses on the distribution of innovation costs and benefits. In order to encourage the production of OIC goods through the use of decentralized networks, OICs emphasize mechanisms that will distribute both the costs and benefits of production across the network; this is the foundation of most OICs' ongoing efforts to take advantage of positive network effects to fuel further innovation and production. 119 In order for this distribution to happen efficiently, technological production, distribution, use, and innovation in OICs must scale easily as the network grows. Both Principle One and Principle Two are foundational to creating these positive effects. Taken together, freedom to operate, freedom to innovate, access to knowledge, transparent and non-discriminatory legal instruments, and interoperability and technology neutrality mean that every participant can both reap the benefits of each new production and contribute indi- 118. See The Open Source Definition (Annotated), supra note 91, § 10; see also id. § 6 ("No Discrimination Against Fields of Endeavor"); id. § 8 ("License Must Not Be Specific to a Product").
119. Positive network effects are economic externalities that benefit every other component of a network with each additional option or input to the network. A common example is the telephone network. Every time an additional phone is added to the network, the network is more valuable to each user because she now has one more resource she can access via the network. In the OIC context, innovation and production can occur ever more efficiently as more innovators enter a project and contribute to it. Both freedom and openness operate to lower the barriers to entry for new innovators, generating positive network effects.

No. 1]
Protecting Open Innovation 25 vidually to any task. The network also benefits as a whole from the distribution of costs among the participants.

D. Principle Four: Reliability
Ensuring the reliability of the normative commitments to freedom, openness, and distributed benefits is another key OIC principle, especially in communities that operate on volunteer or other nonmonetary bases. Reliability arises from both the community's informal policing of norms and from the intellectual property licensing structures that enforce key requirements for the communities. Together, these enforcement structures help convince OIC members to participate in open innovation processes by building in trust and accountability, providing assurances that their contributions will remain subject to the normative commitments that the OIC values. 120 Specifically, there appear to be at least two separate characteristics of reliability critical for OIC functionality: longevity and robustness.

i. Longevity
The first characteristic of reliability is longevity. Simply put, the normative commitments need to last. If they were to expire or be abandoned after only a short period, their associated benefits to the community could be severely limited, both in terms of their intrinsic value and their potential for scalability. The use of standardized intellectual property licenses to enforce norms accomplishes this through the term of the exclusive rights at issue. In most countries, copyright entitlements last at least fifty years and often far longer; patent terms, at twenty years, are shorter but are similar with regard to their ability to enforce norms over a sustained period of time. 121 Thus, intellectual property can serve a positive role in creating longevity if it can be employed as a norm enforcer instead of a norm adversary.

ii. Robustness
The second characteristic of reliability is robustness -the ability of the open innovation regime to withstand challenges and attempts to "game" the system. OICs vary in their tolerance for different uses of However, all desire to avoid the subversion of the community's normative principles and distributed organization. 125 With regard to patents, norm-undermining gamesmanship by patent holders is a significant concern for OICs. 126 In the copyright context, licensing requirements -particularly copyleft-style requirerequirements -are used by OICs to reduce gamesmanship. As we describe further in Part IV, we think the licensing methodologies used to enhance reliability in the copyright context can be applied to OIC patent strategies as well.

B. Existing OIC Patent Defense Strategies
Having examined the theory, values, and practices of OICs regarding copyrights, we now turn to how they have approached the issue of patents. OICs have been concerned about patent threats for some years, and various existing proposals and methods for dealing with problematic patents presently exist. The most prominent strategies include: (1) defensively publishing information to serve as invalidating prior art against subsequent patent applications on the same technology, 127 (2) encouraging patent pledges in which patent holders promise not to enforce against OIC actors, 128 (3) inserting so-called "patent peace provisions" into open source licenses that obligate li-122. We are loosely defining the "open source" community as those who are opposed to the free software insistence on "copyleft" requirements.
123. For leaders in the open source community, the ability to internalize such benefits is seen as an important component of freedom. See Perens, supra note 89, at 183 (discussing the ability to take modifications private under the X license

No. 1]
Protecting Open Innovation 27 censees to avoid patent conflicts within code contributors, 129 and (4) forming patent protection "watchdog" groups that collect pools of patents to assert against offensive threats to the community. 130 In this Section, we examine these strategies in order to extrapolate further principles and means to draw upon in evaluating and optimizing defensive patent strategies.

Defensive Publication
Defensive publication, rather than focusing on obtaining patents for direct defensive use against lawsuits, instead focuses on creating a bulwark against future patent threats by publishing information about OIC technologies in order to create prior art against future patent grants that would cover the technology. This works because, when knowledge is publicly disclosed under the patent laws, no future patent application generally can be granted as to the knowledge contained within the prior art or any obvious improvement thereupon. 131 Defensive publication takes advantage of many of the underlying principles that OICs already employ. For example, because one of the core commitments of openness is access to knowledge, publishing technical information within OICs is generally prolific and consistent with this value. It is also generally low-cost and distributed, as many OICs use public repositories to store their code and communications. 132 The benefits are also distributed, as preventing patents protects all OIC members from threats. It also reflects OIC cultural and political expectations by strongly promoting access to knowledge, freedom to operate, and freedom to innovate. And because the information is in the public domain, defensive publication promotes technology neutrality and interoperability.
Defensive publication has, however, some important limitations. As a practical matter, merely publishing the information is necessary but not sufficient for it to serve as prior art. In order for information to be effective as prior art, it must be publicly accessible and patent examiners and courts must learn about it. 133  for prior art to help lower the search, information, and transaction costs in finding published information for both the USPTO and patent litigation defense lawyers. 134 In addition, despite the theoretically low cost of publishing information online, the majority of defensive publishing efforts have failed to provide scalable patent defense for OICs. We posit that this is for two reasons. First, while OICs generate prolific amounts of technical information, they do not automatically organize their information in ways that speak to the needs of patent examiners, judges, or juries; instead (and understandably), they organize it in ways that speak to engineers. So, for example, an OIC code base will include the code and occasionally documentation or commentary meant for engineers to understand how to implement and fix it, what bugs are currently within it, and what new features might be in the offing. 135 The basics of how the code is designed and how it operates will likely be assumed or glossed over because the code works. Yet this is exactly the information that patent examiners, judges, and juries need to reject claims or invalidate patents. They need basic explanations that describe the broad functionality and potential (and even hypothetical) applications of the technology. They also often need the information described in a manner they can understand if they are not trained in the specific technology or discipline of the publisher.
Second, because patent attorneys are well aware that defensive publication could thwart their clients' desire for patents, they will often attempt to characterize patent claims in words other than those a traditional engineer or innovator in the field would use. It is black letter patent law that a patentee may be "his own lexicographer" 136 and thus use any words he wants -even made-up words that he invents -to describe his invention. So instead of describing a children's slide as a "slide," he may describe it as "an elongated trough" with a "substantially straight traverse cross-section." 137 This is commonplace as a strategy to avoid prior art, and makes it difficult for engineers and other non-lawyers to develop effective prior art data-

No. 1]
Protecting Open Innovation 29 bases that rely on Boolean or linguistic searches to match art to patent claims. Moreover, when the prior art is merely the distribution of OIC source code, searching the code for prior art would be especially difficult. Thus, while defensive publication provides possibly the most distributed benefits of any OIC patent strategy -i.e., the entire world benefits from access to the knowledge published and its ability to invalidate any patent claim that attempts to cover the published technology -the search and information costs associated with finding the right defensive publication limit the benefits of the strategy as a practical matter. 138 Therefore, it is a partial strategy. A more optimal patent strategy would provide the same benefits to all OICs and at the same time attempt to lower costs, especially in the search, information, and transactional areas.
In terms of longevity and robustness, defensive publication also presents some challenges. Longevity requires that the information be available and accessible when needed to defend against a patent application or claim. Therefore, beyond the cost of producing the information, it must be organized and maintained in a manner that reduces the search, information, and transaction costs of decision-makers and advocates. Because patent examiners have, on average, only eight to eighteen hours to complete the review work for each patent, 139 if they do not find the most relevant art immediately, it may not become a factor in the examination. Finally, current public search engines such as Google often discount older information in favor of new, creating further questions about the effectiveness of many defensive publications as the search costs required to find them go up over time.
With regard to robustness, the ability to "game" defensive publication varies, depending on how well defensive publications can be mapped to subsequent patent claims intended to cover the same invention. Defensive publications face at least two challenges to their effectiveness here. First, as noted above, there is an inherent linguistic problem in predictably mapping publications written for engineers to patents written for examiners, judges, and juries. Second, patent attorneys are skilled at playing linguistic games with claim language that make the claims appear different from the invention disclosed in relevant publications. Prior art in the form of patents, on the other hand, suffers less from this problem because patents are generally written for the same audience and often describe inventions broadly, explicitly, and with numerous examples.

Harvard Journal of Law & Technology
[Vol. 26

Patent Pledges
Patent pledges are promises by patent holders not to enforce their patents under certain conditions. In legal terms, the mechanism by which this takes effect is either estoppel (for those who can show reliance on the pledge) or implied license (for those who lawfully acquire and utilize the technology covered by the patents). Among OICs, Red Hat has provided one of the boldest and most prominent examples of such a pledge, promising not to assert any Red Hat patent claim against any use of FOSS. 140 IBM, Nike, the BioBricks Foundation, and others have made similar pledges for various OICs. 141 To date, no such pledge has been tested in court.
Patent pledges appear to offer their benefits in a distributed manner to a particular OIC population, for example, Red Hat's offer is to the FOSS community. However, for any individual member of an OIC to receive an explicit legal benefit from a pledge, she must demonstrate some reliance on, or implied acceptance of, the pledge. Thus, it is unclear how broadly these benefits extend because there is no known record of who has specifically relied on each pledge. In addition, pledges require no reciprocal agreement to keep technology open -for example, a company may rely on a pledge to produce software but make no pledge of its own as to patents that cover that software, reintroducing the patent liability risk to downstream innovators.
Beyond these limitations, there is also the problem of pledge size and scope. While the Red Hat pledge makes non-enforcement promises as to all patents, it does so only as to the use of FOSS, leaving aside, for example, hardware that might be offered under an open license yet run on proprietary software. 142 143 This has led to some uncertainty about the pledge's scope, and concerns that IBM's pledge includes meaningless "junk" patents instead of the more valuable "crown jewel" patents that might pose a more direct threat of litigation. The worry is that IBM may control via patent the most marketable or otherwise valuable aspects of Linux, which it cannot control via copyright due to the GPL's copyleft terms. 144 Similar critiques can be made about other pledges, notably the Eco-Patent Commons and the Creative Commons Green-Xchange approaches, which also provide pledges or licenses only on a patent-by-patent basis, leaving out some important portfolios and leaving the supposed beneficiaries with extensive search and information costs if they want to make sure that they have freedom to operate. 145 This fails to maximize benefits under the criteria of technology neutrality and interoperability. Patent pledges -while they theoretically distribute benefits to anyone who chooses to rely on them -are also limited in their ability to fully distribute costs and benefits. First, the pledge model concentrates the costs within specific companies that already have significant patent portfolios without requiring any of the beneficiaries of the pledges to bear similar costs. They are not distributed in the same way that knowledge, technology, and licenses are distributed in OICs more generally. Second, because there is no reciprocal or viral component to the pledge, this strategy does not take advantage of positive network effects to extend the benefits and costs within the network or encourage those outside OICs to join and grow the network. By limiting the pledges to specific patents and specific OICs rather than covering open technologies more generally, firms that use this strategy fail to distribute both the costs and benefits as widely as possible.
In many instances, patent pledges represent the attempts of large actors, such as IBM, to meet OIC cultural expectations and allay fears of patent threats from companies that see a strategic value to entering the OIC space but have not previously used open innovation as a model. In practical terms, patent pledges do provide some assurance of freedom and openness. At the same time, as noted above, they are limited in terms of their ability to support expectations of technology neutrality and interoperability by the common feature of restricting

Harvard Journal of Law & Technology
[Vol. 26 the protection to specific OICs. Further, even for pledges offered by a member of an OIC, there has been some tension as to the exact commitment embedded within each pledge, especially as there is no standardized legal language for OIC pledges. 146 Beyond their limited reach in practice, a lack of perceived reliability is likely the greatest limitation on pledges' ability to alleviate patent concerns in OICs. Patent pledges are generally quite transparent in their terms. Further, in theory, they last for the full duration of the relevant patent terms, so the longevity criterion seems to be satisfied. However, their level of robustness raises concerns. First, because pledges rely on the legal doctrines of estoppel and implied license, they are susceptible to attack while they are in effect. Estoppel is a fact-specific defense to infringement in which an alleged infringer must show that she knew of the patentee's promise and then materially and reasonably relied upon it. 147 Implied license requires a strong showing of the patentee's intent to license for a specific use (as opposed to a more general set of permissions); further, the license can be withdrawn under many circumstances, such as when there is no explicit consideration in return for the product. 148 Therefore, reliance on an extant patent pledge carries uncertainty.
Second, patent pledges are entirely under the control of the pledging entity, which can choose to withdraw them at any time, leaving stranded anyone who cannot already show reliance or active, specific uses. Therefore, an entity that pledges important patents can, in theory, strategically withdraw its pledge once a technology becomes central to a network or otherwise especially valuable. 149 And even if a pledging firm remains committed to its initial strategic decision to pledge patents, the pledged patents may later be sold to entities, including trolls, that did not make similar promises. 150 Given these limitations, it is unclear that pledges will produce broad-based reliability for OICs with regard to patent threats. However, harnessing the benefits of pledges -especially their transparency and inexpensive applicability to all who are engaged in innovation 146 149. This likelihood is no doubt limited by the damage a firm would incur to its reputation by taking such an action. Further, to some extent, OICs can do what they do bestdesign around the issue -but this response is limited in effectiveness when the patents in question are sufficiently broad. It is a more effective response in the copyright realm, where threatened code can be rewritten to overcome the more narrowly construed copyright protection.
150. Bankruptcy is a particular concern. See supra notes 64-66 and accompanying text.

No. 1]
Protecting Open Innovation 33 covered by the pledge -to a strategy that provides more reliability would increase robustness. As we will discuss in Part IV, this can be accomplished through more explicit assent, commitment, and accountability mechanisms that rely on legal doctrines with lower burdens on OIC users, obligations to enforce the commitment as to successors-in-interest, and limiting the ability of pledging firms to engage in gamesmanship by offering only some relevant patents under the pledge.

Patent "Peace" Provisions
Patent "peace" provisions are specific terms in OIC licenses that require any licensee of the OIC technology to forgo patent lawsuits against the licensor and, in some forms, other licensees or users of the same OIC technology. The most famous example of the latter is Section 10 of Version 3.0 of the GPL ("GPLv3"), which terminates the GPL's copyright license, along with any patent licenses, in the event that a user initiates a patent lawsuit against a contributor to any GPLv3 licensed program. 151 Peace provisions are structured as legally binding commitments, either as a general covenant not to sue or as a specific license to a particular OIC technology. 152 Other examples include the Apache license 153 and the Yahoo! DomainKeys license. 154 Patent peace provisions provide the distributed benefit of safety from patent threats from known OIC licensees, which promote access to knowledge of the OIC technology and freedom to operate and innovate. The scope of the benefit varies based on the size and power of licensee patent portfolios and whether the covenant not to sue covers just the licensor or all users of the OIC technology. The costs are similarly distributed within the network of license users, although one might assume that the relevant costs of patent acquisition and enforcement are concentrated among the larger entities with internal patent-related legal resources.
However, much like the patent pledge model, there is a limit to the scale of this model. Because the benefits are limited to users of the license or technology at issue, patent peace provisions do not provide any structures or incentives to grow the network and maximize the positive network effects of the defensive practice. Nor do they emphasize technology neutrality or interoperability with other OIC technol- ogies. Thus, the costs and benefits, while distributed to various extents, are still relatively limited in both scope and potential. Patent peace provisions also meet cultural and political expectations in similar ways to patent pledges, especially when they emphasize freedom from patent limitations and are transparent in their promises and approach to the legal issue. However, again, because of the limited network effects, technology non-neutrality, and incentives to contribute to collective patent defense, peace provisions are limited in substantial ways. If anything, these provisions discourage proprietary communities from using OIC technologies for fear that they will forgo any value to their patents without significant benefits. The provisions also lack some transparency in that they do not provide a clear mechanism for identifying the patents that are specifically subject to the "peace" protection.
In terms of longevity, patent peace provisions last for the full patent term. However, no patent peace provision has ever been tested in court, so like other approaches, it remains to be seen how robust they are. One can imagine some challenges to the validity of such an agreement, especially for later-acquired patents or patents that have been sold off to third parties who then claim there is no privity with the original licensee.

Patent Protection Groups: The Open Invention Network Model
Patent protection groups are consortia or organizations that gather patents in order to pool their resources for a single purpose. One version of this is a patent pool, which can be an effective mechanism for organizing intellectual property licenses for technical standards 155 and as a means of aggregating enforcement. The Open Invention Network ("OIN") has taken this approach to patent defense 156 for the Linux operating system. 157 Because the organization and approach can vary from patent protection group to patent protection group, we use OIN as an example of a particularly open and thoughtful version of the "pool" approach to OIC patent defense. 157. We refer to "Linux" as an operating system for readability purposes, but we acknowledge that this is shorthand. See Richard Stallman, Linux and the GNU System, GNU OPERATING SYS., http://www.gnu.org/gnu/linux-and-gnu.html (last updated Oct. 25, 2012) (explaining the difference between the GNU operating system and the Linux kernel, and the combination of the two as a "GNU/Linux" system). If Linux is threatened by a patent lawsuit, OIN stands ready to assert the patents in its pool to neutralize or diminish the threat. OIN has received fiscal or intellectual property donations from IBM, Novell,Philips,Red Hat,and Sony. 159 In terms of costs and benefits, OIN concentrates its costs internally and among its fiscal and intellectual property benefactors. It distributes the benefits of its defensive portfolio to anyone who takes a license. And because OIN is committed to protecting Linux from patent threats, 160 the defensive benefits of its patent pool are, at least theoretically, distributed further to anyone who uses Linux or innovates on top of it. However, OIN's centralized, hub-and-spoke model, in which the OIN hub engages in reciprocal licensing with many licensees using Linux, trades a strong, centralized ability to enforce patents in defense of one important product (Linux), for a more general distribution of benefits. So, for example, OIN provides few direct benefits to other OICs, including the Apache web server, Android, or FreeBSD, as well as the many applications that run on Linux. This leaves other OICs vulnerable to patent attacks even if they use Linux. For this reason, OIN's benefits are also limited in their network effects by the lack of technology neutrality and interoperability. 161 The OIN model adopts many of the principles and mechanisms of OICs. For example, OIN creates transparency by openly providing a list of all the patents it owns, a copy of its standard license to its pa- 161. Its costs are also somewhat obfuscated but seem concentrated among the firms that support it. While this may be sufficient to meet the cost-benefit ratio required for Linux as an OIC, it does not appear to extend to other OICs.

Harvard Journal of Law & Technology
[Vol. 26 tent portfolio, 162 and a list of its licensees. 163 The license is available to any entity that chooses to agree to its terms (which, importantly, include a license back to OIN of any patents that "read on" Linux), and is thus non-discriminatory. 164 By defending Linux -the kernel of a common FOSS operating system and thus a central component of many FOSS production models -OIN supports the OIC norms of freedom and openness, 165 including freedom to operate and innovate, access to knowledge, and non-discrimination.
On the issue of reliability, OIN's patent license seems to create longevity over the term of each patent committed, and seems to be at least as robust as other OIC licenses by remaining in place unless and until the licensee asserts patents against Linux. However, the reliability of models that depend on a centralized organization is less clear. The strength gained by centralizing costs and administration into one organization with the ability to act on behalf of the community requires both organizational longevity and a robust commitment to its stated defensive positions. Not unlike Google's pledge "Don't be evil," 166 OIN has promised to be good, and to defend Linux according to the community's values. However, it is unclear if this is a legally binding commitment and if mechanisms for transparency or accountability exist to ensure or enforce it. Thus the OIN model relies heavily on trust and the value of its brand. However, there is no legal guarantee that OIN will continue to be a good actor. It could potentially sell off its patents to an offensive actor in the future that could target non-Linux OICs. This seems unlikely for OIN specifically, 167 but other centralized pools may have less robust ties to the OIC community. More quotidian organizational issues -such as lack of funding or lack of leadership -can also arise, leaving stranded the community relying on a troubled patent protection group. This is a risk exacerbated by the relatively concentrated costs that arise if OICs shift to new technologies, standards, or platforms beyond those previously defended, i.e., if users move from Linux to FreeBSD. 165. However, it is worth noting that because of its focus and centralization, OIN's transparency with regard to its actual enforcement strategies and actions is much more limited in relation to its openness in disclosing all patents, licensees, etc.
167. At the same time, whether OIN is fully trustworthy is the subject of some controversy in the FOSS community. See Dana Blankenhorn, Mueller Calls OIN a Scam, ZDNET (June 2, 2010, 5:26 PM), http://www.zdnet.com/blog/open-source/mueller-calls-oin-ascam/6613. Regardless of OIN's actual trustworthiness, any organization that must rely on trust in this way may generate similar controversy.

No. 1]
Protecting Open Innovation 37 tuting a defensive pool or organization on a technology-by-technology basis could be substantial and prohibitive.

C. Synthesizing a Defensive Patent Strategy for OICs
Each of the above strategies has some merit, though each has important limitations in its reach and effect. We recognize that some more limited solutions to OIC patent threats are intended to solve specific problems, and that each makes tradeoffs that may maximize the benefits demanded by these specific situations. Yet we think it is possible to develop a complementary approach that is more broadly applicable and that has the capacity to spread the benefits of defensive patenting across OICs. Drawing on our analysis of defensive patenting, OIC literature and practice, and the reach and limitations of the existing approaches we have discussed, we synthesize a framework with which to evaluate OIC patent defense strategies and to develop a more generalized approach. The framework addresses each of the barriers to defensive patenting we identified in Part II, and encompasses the four principles we distilled in Part III.A. Synthesizing these analyses shows that, to be most effective within the OIC context, a defensive patent strategy should maximize the following characteristics:

Distribution of Costs and Benefits
• With this framework in hand, we now turn to describing and evaluating our proposed Defensive Patent License strategy.

IV. THE DEFENSIVE PATENT LICENSE: A NEW STRATEGY FOR ADDRESSING PATENT THREATS TO OICS
Each of the above approaches goes some distance in addressing patent threats for OICs, but each has limitations in its reach and effect. Therefore, working from our framework, we propose a new model for [Vol. 26 addressing OIC patent threats that is designed to maximize important benefits of existing models and address some of their limitations. Specifically, we model a distributed cross-licensing system anchored by an OIC-style standardized open patent license -the Defensive Patent License, or DPL. The DPL blends the general strategy of defensive patenting with the OIC values of openness and freedom, using a FOSS-inspired, decentralized and standardized intellectual property license to distribute costs and benefits, and to provide an interoperable, technologically neutral, reliable, and legally binding commitment to defense.
The DPL draws upon and reflects three basic principles, articulated for us by Brewster Kahle, founder of the Internet Archive, during our initial conversations with OICs about this idea.
The DPL is a license that reflects a public commitment to defense, so our patents are forever defensive.
Any organization may freely use these licensed patents while also being publicly committed to defense.

An organization that commits a breach of this license is ceasing its commitment to defense.
Overall, we believe that this approach could significantly lower the barriers and maximize the benefits of defensive patenting for OICs. In this Part, we describe the DPL and explain how it can help OICs protect themselves against patent threats.

A. The DPL Model
The DPL is a standardized open patent license designed to encourage the creation of a broad, decentralized network of OICs that both patent their innovations with a commitment to defensive purposes and license them on a royalty-free basis to any others who will do the same. 168 In doing so, the goal is to build up a collective network of patents that has the same deterrent power as a large proprietary defensive "portfolio" -but a portfolio that has its costs and benefits distributed across the users of the DPL, and that does not require separate centralized management. 169 As such, we harness the network effects of OIC distributed cost and benefit structures, the commitment to OIC 168. For illustrative purposes, we provide a model version of the DPL in the Appendix to this Article.
169. We do suggest a centralized information management system for tracking DPL transactions but not for acquisition or enforcement, which constitute the largest costs of centralized defensive patenting consortia.

Protecting Open Innovation 39
values, and the reliability of other standard distributed OIC licenses, such as the GPL or Creative Commons' copyleft licenses, and apply them in the patent context. The DPL operates by creating a set of viral, bilateral obligations focused on preventing offensive patent litigation and promoting freedom to operate and innovate. Specifically, the DPL provides every DPL user a perpetual, worldwide, royalty-free license to every other DPL user's entire current and future patent portfolio, 170 subject to the following four conditions: 1. Every DPL user (licensor or licensee) will forgo any offensive patent infringement actions against any other DPL user; 2. Subject to Condition 4, every DPL user will offer her entire current and future patent portfolio under the DPL; 171 3. Every DPL user will bind any successor-ininterest to any part of her patent portfolio to her obligations under the DPL; 172 and 4. If a DPL user wishes to stop offering her patents under the DPL, she may do so, but only with six months' notice to existing DPL users and future parties. She must continue to grant, and may not revoke, any licenses that are in place before the end of the notice period. 173 Once she stops offering the DPL, 170. As defined, a patent portfolio includes every issued patent as well as any continuations, divisionals, applications, or provisionals. In other words, any filing with any patent office in the world that lays claim to a patent right of any kind is subject to the DPL. For users without any patents, they must simply commit to putting any patent rights they do acquire in the future under the DPL.
171. Much like Creative Commons, we have also considered the possibility of a suite of DPLs which could offer variations, both in terms of the level of portfolio commitment (for example, the whole portfolio versus a defined portion) as well as the strength of the licensee's viral obligation (for example, only a promise not to sue versus a promise to make all licensees agree to the same terms). While we see the merit to these variations, we will describe just the basic license in this Section and then discuss the pros and cons of the variations below.
172. There is one exception to this obligation: due to the common practice of having employees automatically assign patent rights in their inventions to their employers, patents that are subject to such agreements are exempt from the obligation to be licensed under the DPL unless the employer assignee is already a DPL participant. Otherwise, the inventor would breach her personal licenses every time the automatic assignment took place. This situation can also be addressed via the "inventor scope limitation" discussed below.
173. After this period, previously issued licenses remain in effect, but there is no obligation to issue licenses to new parties or on newly filed patent rights. Six months gives DPL

Protecting Open Innovation 41
3. Once registered, the website provides the inventor with a blanket DPL licensing form that she agrees will bind her in relation to anyone who accepts the license; 4. She then receives access to the pages for every other DPL user with an option to accept any or all of the DPLs for their portfolios. One can also imagine an option to accept all known DPLs in a single click. The website then distributes an acknowledgment of all license offers and acceptances to the participating parties and records them in its internal database. The licenses take effect immediately.
5. As new patents issue or new applications are published for each registered DPL user, they are added to the website, either automatically or by the DPL user herself. Each DPL user may then take additional licenses to newly issued patents either manually or automatically via the DPL website.
6. If the DPL user chooses to stop offering her patents under the DPL, she registers her six-month Discontinuation Notice at the website, which alerts all DPL users. Six months later, the DPL user's page is updated appropriately but all information related to the user and her portfolio remain preserved. Other DPL users are then given the option to revoke any licenses for the leaving user as of the Departure Date.
The figure below captures this sequence of events. The DPL thus provides incentives for innovators to legally commit themselves to defensive use of patents, rewarding participation with the benefits of unfettered access to the patents of other DPL users and freedom to operate with respect to DPL technologies. As the number of DPL users grows, they develop into an extensive cross-license network, with positive network effects eventually arising, increasing the benefit of joining and causing the network to grow more rapidly. The DPL provides disincentives to leave the network, as those leaving may potentially lose their licenses while those who remain keep theirs, even those licenses from the leaving party. In this way, patents subject to the DPL remain reliably defensive for all DPL users who remain in the network. For those outside the network, all DPL-licensed patents are still active threats; DPL users may seek licenses or enforcement against outsiders on any terms they wish.

Distribution of Costs and Benefits
We begin by evaluating the DPL under the first prong of our framework, which suggests that distributing the costs and benefits of patenting is likely to be an important factor in any defensive patent strategy for OICs. By leveraging the distributed open license model common to many OICs, the DPL tackles the problem of concentrated licensing and information costs and benefits. Because patent licenses are required to be equally and freely available to any DPL user, the benefits of each additional patent licensed under the DPL can be distributed equally and easily among the licensees. Through the same mechanism, the costs of patenting are also distributed, albeit not quite as equally. Much like the code production in FOSS projects is distributed, with some coders contributing proportionally more than others, but with any coder able to participate as her resources allow, so would innovation and patent prosecution be distributed in the DPL network.
By actively distributing costs and benefits in this way, the DPL increases access to patented technology, expanding the supply of tools for innovation. For example, if a member of an OIC wishes to find a particular technology to use and wants to make sure it comes with the freedoms that OICs support, he can search the DPL website for the appropriate patents and immediately receive licenses to them under the DPL. And much like other open licenses and formalized intellectual property pooling efforts, the DPL facilitates earlier and more widespread use of patented technology via non-discriminatory and royalty-free licensing and more rapid advancement of technology by clearing blocking positions and permitting more firms to improve on existing patented inventions. By encouraging a global network of patents licensed under the DPL, the quality of DPL-licensed products will improve more rapidly than if DPL users continued to be limited in terms of the patented technologies they could access. This is particularly true for firms with limited resources.
With regard to enforcement, our model does not include a centralized "brain" to coordinate enforcement actions, which makes it far less costly to administer and, we think, more reliable over time than a centralized model requiring staff and substantial ongoing resources. While the lack of a centralized brain also means that there is no centralized responsibility for enforcement, DPL users who share similar OIC values or who wish to come to the aid of other users who are under attack can either file protective lawsuits directly or offer to pay the costs of suit in return for initiating litigation. 176 This model therefore lacks a guarantee that enforcement efforts will occur, but also allows maximum flexibility in how each DPL user chooses to enforce their patents in specific situations. While the specific circumstances of each suit are too difficult to anticipate, we believe there will be enough common interest and common ground for defensive efforts to occur, and for the network to represent a credible deterrent against many offensive patent threats.

A. Distributing Costs
By allowing OICs to easily cross-license each other's patents, the DPL reduces the overall cost of patenting. This is not to say that all firms are likely to contribute equally: firms with limited resources may, at least at first, mostly benefit from access to others' patents. On the other hand, firms that already patent for other reasons, such as to attract venture capital or in order to license or enforce against non-DPL parties, would automatically contribute their patents to the DPL at no additional cost. So long as the overall network contains sufficient patents, the benefits of gaining licenses are likely to encourage additional users to begin using the DPL; as users gain patents, the network's collective portfolio will grow accordingly.
Moreover, we anticipate that, with a binding commitment to use the DPL, some OICs will choose to patent innovations they otherwise might have dedicated to the public domain in order to strengthen the defensive power of the DPL network. 177 We think this is particularly promising among OICs with strong user innovation communities. And this appears to be a broader trend; according to Baldwin and von Hippel, user innovation in OICs is growing. 178 Yet very few of these users patent their innovations. If there were a robust patent defense strategy that benefited from such patents, such as the DPL, these innovators might be convinced to patent their contributions and thus strengthen the network even more. This would also give them recognition for ("[A] right to sue clause cannot negate the requirement that, for co-plaintiff standing, a licensee must have beneficial ownership of some of the patentee's proprietary rights.").
177. We also hope to encourage pro bono patent prosecution for firms that commit to the DPL as an additional way of lowering the barriers to building the network. For example, the USPTO has recently approved a pilot program for legal clinics at U.S. law schools to assist in patent prosecution for parties who could not otherwise afford legal representation before the USPTO. In addition, the recent America Invents Act includes a provision allowing large fee discounts (up to 75%) for "micro entities.  User and Open Collaborative Innovation (Harv. Bus. Sch. Working Knowledge, Paper No. 10-038, 2009), available at http://hbswk.hbs.edu/item/6325.html.

No. 1]
Protecting Open Innovation 47 their work more broadly, and even potentially improve their value in various employment markets. Though the basic costs of patenting -filing fees and attorneys' fees -remain the same for each individual patent that a DPL user might pursue, 179 we think the DPL would greatly reduce important transaction costs related to defensive patenting. First, it facilitates easy cross-licensing by eliminating the need for costly negotiations between DPL patent holders. Second, it lowers information costs by providing a consistent baseline license with predictable terms for every user. Third, by providing a centralized website with all relevant information and an easy means of accepting DPLs, it greatly limits the legal and search costs associated with assessing the validity of potential patent threats from DPL-licensed patents.

B. Distributing Benefits
As noted above, by making patents available, neutrally and openly, to anyone who also agrees to use the DPL, the DPL distributes the benefits of patenting across the entire DPL network. Importantly, this distribution occurs across myriad OICs -any OIC, regardless of sector, that chooses to use the DPL -instead of across only individual communities or technologies. Anyone in the network can make, use, or sell any technology licensed under the DPL, making it technology neutral, scalable (as each new user simply becomes part of the network and does not impose significant costs on any central organizing entity), and interoperable. Further, anyone in the network can rely on the defensiveness of each of the patents under the DPL with little to no cost. 180 Defensiveness is distributed across the network by virtue of the fact that licenses are irrevocable unless a licensee asserts her patents offensively against another DPL user or leaves the network. 181 This allows each DPL user access to a networked defensive patent portfolio potentially far greater than she would have been able to ac-179. Note, however, that these costs might be reduced via pro bono patent prosecution and common patent fee funds for innovators who commit to putting any patents they acquired under the DPL.
180. Those familiar with open copyright licenses such as Creative Commons and the GPL will notice that the scope of the DPL is much broader. Rather than merely applying to future derivatives of the originally-licensed code, it applies to all licensee patents -past, present, and future -unless and until the licensee leaves the network. This is intentional, as patents are not congruent with copyright, and have much more damaging potential, precisely because they apply beyond derivatives, to anything the patent reads on. They claim the entire space. So in order to achieve defensiveness, the license has to address the scope of the threat. Moreover, the notion of a "derivative" in copyright licenses rests on the statutory definition which patent law lacks. See supra Part III.A.2.A.i. 181. The GPLv3 uses a similar mechanism in its "patent peace provision," which terminates the copyright license, along with any patent licenses, in the event that a user initiates a patent claim against a contributor to the licensed program. See The GNU General Public License v3.0,supra note 22. [Vol. 26 quire alone, and creates a potential arsenal that should greatly deter offensive users of patents. 182 As the network grows, the defensive benefits grow accordingly and are distributed across the network. 183 Thus, as the DPL gains adoption, the number of DPL users will grow and the network effects of the license will increase the benefits to each licensee. This in turn will increase freedom to operate and the overall commitment to defense for those in the network, while simultaneously offering increasing incentives to those outside the network to join. It also offers the benefits of establishing "gold standard" prior art that prevents non-OICs from patenting OIC technology. The potential to assert OIC patents against non-OIC entities for damages, royalties, and injunctions, is an incentive to join the network, since non-OICs are not covered by the definition of defensiveness. Such assertions could also occur as a matter of community defense when a non-DPL entity threatens a DPL user without any patents of its own.
As compared to some of the other defensive strategies we have considered here, 184 the DPL can provide many of the same benefits, but distributed more broadly, across any technology sector with a DPL uptake. There is no need to create an entity to manage a patent pool, and no need to evaluate which patents should be put under a defensive pledge. DPL users can still choose to offer pledges or agree to patent peace provisions outside of the DPL network, but may choose to forgo these strategies as the network grows. In addition, because OICs will be patenting defensively instead of publishing defensively, their innovations will become more effective prior art within the USPTO and courtrooms, helping to prevent subsequent patenting of anything they use.

C. Possible Challenges to Generating Network Effects
For the DPL to most effectively distribute both the costs and benefits of defensive patenting across OICs, a sufficiently sized network of DPL users must develop. Therefore, if barriers to adoption are too high, or incentives are too low, then the DPL is less likely to generate the positive network effects that would make it most useful to OICs.

Protecting Open Innovation 49
We address two important concerns about incentives and barriers to adoption here: concerns about barriers created by the DPL's "all-in" requirement, and concerns about sufficient capital investment in OICs.

i. Balancing Barriers Created by the All-In Requirement
A notable feature of our model DPL is its all-in requirement that firms offer all of their patents under the DPL rather than picking and choosing individual patents from their portfolio. This could represent a barrier to entry for both traditional proprietary firms and OIC firms. In our conversations with representatives of both types of companies, individuals suggested that they might well be willing to put some of their patents under the DPL, but asking them to put all of their patents under the DPL would be too large a commitment. In particular, where companies had split business models for different products, the idea of having to license one's entire portfolio presented a challenge: for example, proprietary versus OIC, or where their core businesses were ultracompetitive (and thus less amenable to broad licensing), while more peripheral parts of the business were less competitive.
At the same time, a variety of entities are concerned about gamesmanship -and the all-in model limits this concern substantially. Allowing DPL users to make only partial commitments would open up extensive opportunities for gamesmanship by offering only select (perhaps "junk") patents, and would increase the information costs of assessing which patents of a firm are valuable, which are necessary, and which are potential threats. If all patents are licensed, by contrast, there are far fewer determinations to be made; one simply takes a license to all desired patents under the DPL for free. Therefore, gamesmanship concerns must be weighed against worries over closing off possible future business models or limiting a firm's ability to split existing models. In the aggregate, we think that the all-in model creates the most benefit in this balance, and that the cost for many OIC firms is relatively low. Nonetheless, we contemplate three possible variants on the scope of the DPL that would allow for partial portfolio commitments while limiting gamesmanship: (1) where the DPL user commits all patents that are defined under a specific independent technical standard, (2) where the DPL user commits all patents that are invented by specific inventors, and (3) where the DPL user commits all patents that cover a technology as defined by a neutral third party, such as a regulatory agency (for example, the Federal Communications Commission). In these instances, the reciprocal scope of the licensees' DPLs would only cover the technologies that substantially embodied these standards, the work of these inventors, or the technologies that were the subject of these regulations, respectively. Using such neutral, objective, and verifiable categories to define the scope of the DPL commitment would limit the amount of gamesmanship one could engage in and force the patent holders and licensees to share the information cost of determining which patents fall under the license. So, for example, a company such as Facebook could offer DPLs on all of its patents that list Mark Zuckerberg as an inventor. This would limit the number of patents under the DPL, but it would also limit the scope of the reciprocal licenses to Facebook. Facebook would only receive reciprocal licenses to projects that substantially embody Zuckerberg's personal work. Projects he has not worked on could still potentially infringe. Thus, the incentives would be to add more inventors to the list of "By Inventor" DPL patents in order to provide more freedom to operate for more projects. And by leaving out patents by other inventors, it would signal to the community that those patents are more likely to be used offensively. It would also allow specific inventors to bargain in their recruitment or employment contracts to have patents they invent put under the DPL without forcing the company to put its entire portfolio under the license. Therefore, while our Model DPL is based on the all-in model, these other variants are also possible.
Another variant worth discussing is what David Hayes and Eric Schulman call the "non-sticky" option in their response to this Article. 185 The Hayes and Schulman approach preserves the entire portfolio aspect of our all-in model while initially limiting exposure to future regrets from issuing irrevocable licenses. 186 Under this option, the DPL licensor would publicly announce a specific time period (for example, between zero and sixty months) where their participation in the network would be "non-sticky" -meaning that it could give its six-month notice and then abandon any DPLs it undertook, revoke any DPLs it issued, and leave the network. However, if the announced time period elapses and the licensor is still part of the DPL network, 185. Hayes and Schulman have discussed their thoughts on this option, among other considerations and responses to the DPL. David L. Hayes & C. Eric Schulman, An Early Response to a Recent Proposal for a Defensive Patent License (DPL) 6 (2012) (unpublished working paper), available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2054314. We sincerely appreciate their time and thought in their work on this topic.
186. Hayes and Schulman have also provided a number of additional suggestions and useful technical edits, many of which we have adopted in the Model Defensive Patent License provided in the Appendix to this Article. There are two we have not adopted, but would like to call attention to here. First, Hayes and Schulman note that companies with substantial patent portfolios may be especially worried about competitors creating cloned products or services under the DPL, and suggest carving out cloning and foundry products. See id. at 8, 11. Second, they identify the conundrum that existing exclusive licenses might pose for potential DPL users, and suggest carving out all patents exclusively licensed prior to a user joining the DPL, provided that these licenses were entered into in good faith and not to circumvent the DPL's requirements. See id. at 8. We worry that these good faith requirements would be difficult to enforce effectively, but we do agree that exclusive licenses pose an issue for prospective DPL users.

No. 1]
Protecting Open Innovation 51 its licenses would become "sticky," meaning that they become irrevocable even if the licensor left the network, except under the offensive conditions outlined above in Part III. A. 187 This sticky/non-sticky option strikes us as a smart way to encourage what Hayes and Schulman call "Substantial Portfolio Companies" to participate in the DPL network. 188 For example, a company such as Google might wish to experiment with DPL participation before committing its entire portfolio permanently. In such a scenario, it would offer DPLs to all of its patents but designate the license as "non-sticky" for, say, twenty-four months. This would be part of the license and thus transparent to any licensee. If Google finds participation to be positive, then after twenty-four months, all of its licenses become automatically sticky. However, if Google wishes to revoke its DPLs, it can do so as long as it notifies all licensees before the twenty-fourth month. Hayes and Schulman argue that this approach has multiple advantages. It provides SPCs with incentives to try the DPL, allowing them to "dip their toes" in the DPL pool and gain data and experience in order to make a much more realistic and practical assessment of the benefits and costs. 189 Second, it allows each licensor to choose its own sticky deadline and thus customize the license to its particular business model and community norms. 190 Third, it establishes "stickiness" as the ultimate default so that unless a licensor takes explicit actions to notify licensees, the obligations will become permanent. This not only encourages most licensors to follow the default, but also means that those who do revoke their licenses will stand out more and likely generate press and publicity for doing so, signaling clearly that they are no longer committed to defensiveness and putting the rest of their community on alert.

ii. Concerns about Capital Investment
During our presentation of this idea to members of various innovative communities, some participants have expressed concern that the DPL will be practically unavailable to some OIC firms pursuing venture capital because they have foregone the opportunity to assert their patents offensively against competitors or for licensing revenue, a feature that could undermine incentives for venture financing. First, 187. See id. at 6. 188. See id. at 2. 189. See id. at 7. 190. For OICs in particular, we expect there to be external (from talent prospects and competitors) and internal (from employees) pressure to minimize the non-sticky state. Certainly in terms of labor market dynamics, it would allow companies to distinguish themselves when they recruit engineers. On the other hand, we might also expect pressure from investors, business units within a multi-business company, board members, or others to maximize perceived flexibility, which could militate in favor of maximizing the non-sticky state. Accordingly, we see some risk to this option. [Vol. 26 it is important to recognize that the DPL may in fact not be for every firm. For firms with business models that heavily depend on offensive enforcement of patents for exclusion or licensing revenue, the DPL may not be a viable model. However, it is also true that the DPL does not "disarm" patents against all other entities; rather, it only promises that the DPL user will be defensive as to others who make the same commitment. Those who are not part of the DPL network can be sued for patent infringement or approached for licensing revenues by a firm using the DPL just as easily as by a firm that is not using the DPL. Moreover, in return for licensing its patents under the DPL, a firm gains royalty-free access to all patents held by other DPL users. Both of these significant benefits to the firm's bottom line would need to be weighed against any license revenue from DPL users that would be lost. More generally, the DPL is modeled specifically for OIC firms, for which the enforcement of intellectual property has rarely, if ever, been a primary pathway to profitability. Instead, OICs focus on other incentives for production and profit, using intellectual property to secure freedom and openness, as reflected by the DPL.
Based on our framework analysis, we think that maximizing the distribution of both costs and benefits is an essential component of a successful defensive patenting strategy for OICs. Therefore, our model attempts to maximize these possibilities while avoiding barriers to entry substantial enough to prevent the network from growing sufficiently to allow for effective distribution. Not only does this approach mirror and harness the distributed approach to production that OICs already employ, recreating the success of those models, but it also matches the normative cultural and political preferences of OICs much more closely than traditional patenting strategies.

Respect for Relevant Cultural and Political Values
The second framework criterion -reflecting OICs' cultural and political values -is essential to any patenting strategy aimed at helping them. As described more fully above, many OIC participants have strong political or philosophical concerns about software, Internet, and other abstract patents, and in some cases, about the patent system more generally. The strong exclusive rights offered by patents are seen to work directly against core community values of freedom and openness. 191 Therefore, it is necessary for any patent strategy aimed at helping OICs to both respect and reflect these values. While the most 191. See, e.g., Red Hat Statement, supra note 128 ("Red Hat has consistently taken the position that software patents generally impede innovation in software development and that software patents are inconsistent with open source/free software."); Patent Reform is Not Enough, GNU OPERATING SYS., http://www.gnu.org/philosophy/patent-reform-is-notenough.html (last visited Dec. 22, 2012).

No. 1]
Protecting Open Innovation 53 direct approach might be to restrict patentable subject matter to areas that are less threatening to OICs, in light of Bilski v. Kappos, we think it is unlikely that there will be a successful attack on the appropriateness of software as patent subject matter any time soon. 192 Therefore, as a practical matter, the DPL model turns to private ordering alternatives to diminish the perceived threat of software patents rather than waiting for legal reform on the issue. In this regard, both the software patent debate and the DPL as a method for addressing it mirror the debate over software copyright and the development of FOSS licensing. Originally, many programmers objected to the copyrightability of software, much like the current objections to the patentability of software. 193 However, in light of continuing court decisions and legislation reinforcing software's copyrightability, the FOSS licensing schemes became more accepted in the community as a mechanism to mitigate harms. Given the Bilski decision and the current trends in patent legislation, we believe a case for a similar instrumental approach to software patents makes sense, and have designed the DPL to maximize the norms of freedom and openness, and to use the licensing methodology that is now accepted among OICs for copyright. 194 In doing so, the DPL approaches patents in a manner that is more culturally and politically compatible with OICs than many patent pools, and certainly most individual defensive portfolios. 195 More specifically, the DPL explicitly supports broad freedom to innovate, access to knowledge, and protection from legal constraint. All DPL users must give other DPL users full freedom to operate and innovate vis-à-vis their patents, and must offer their patents transparently, neutrally, and robustly. By making freedom and openness central requirements of the license, the DPL attempts to address the community's understandable lack of trust in defensive patenting by individual entities.  REV. 2308REV. , 2347REV. -56 (1994. 194. This is also somewhat similar to the trajectory of Creative Commons. After Lawrence Lessig's constitutional challenge to the Copyright Term Extension Act was rejected by the Supreme Court in Eldred v. Ashcroft, 537 U.S. 186 (2003), he invested his energy into Creative Commons and its private ordering system as an alternative way to support free culture. See Meera Nair, Eight Years after Eldred, FAIR DUTY (Jan. 14, 2011, 07:18 PM), http://fairduty.wordpress.com/2011/01/14/eight-years-after-eldred.
195. For example, because the DPL's defensiveness, openness, and freedom attributes are requirements of the license, questions of the patent holder's present or future motives are much less likely to arise. Whereas Red Hat must reassure the community of its commitment to defensiveness and general opposition to software patents, see supra note 128, the DPL enforces community norms as inherent terms of its license. [Vol. 26

Ensuring Reliability
Our third framework criterion -reliability -is a main reason why we chose a licensing structure for our approach. Where defensive pledges may change with future decisions by the patent holder, the DPL's general irrevocability and successor obligations are intended to ensure that defensive patents remain defensive and available to any other DPL user. Therefore, any DPL licensee can assume that it will be safe from patent lawsuits asserting any DPL patent continuously over time -alleviating concerns that the licensee's freedom to operate could evaporate in the event that a licensor changes its strategic approach or is acquired. Further, because the license can be revoked if (and only if) a DPL user stops acting defensively against other DPL users, it supports a reliable commitment of defensiveness within the network of DPL users. Therefore, the license structure greatly alleviates OICs' strong concerns that entities that state they are engaging in defensive patenting -yet make no enforceable promises to remain defensive -will change their minds later. If an entity chooses to use the DPL, then the commitment to defensiveness becomes reliable.
Therefore, longevity and robustness are built into the DPL through the license structure and terms. Besides being largely irrevocable, all DPLs are perpetual and fully paid up, so they will last as long as the patent does and expire only at the end of the patent term. Second, in addition to "traveling with the patent" to a new entity if patents are sold, we believe that the DPL can be drafted to survive both bankruptcy and any antitrust complaints. Third, rather than relying on the branding and trust capital of a centralized organization, the DPL self-executes; if a licensee abandons defensiveness, any of its licensors may choose to enforce against it. And while each independent DPL entity makes an independent decision about whether to enforce, the DPL itself (rather than a centralized, non-transparent decision-maker) sets the terms for and definitions of offensive and defensive legal action, giving all parties clear enforcement conditions so they can adjust their behaviors and plan for risks accordingly.
As noted above, a major reliability challenge for any defensive model is gamesmanship. In this case, we attempt to balance the need to protect against gamesmanship with the benefits of low barriers to entry and low administrative costs. The first concern is junk patents: if requirements for using the DPL are not sufficiently robust, then there is a risk that an entity will simply offer patents that are not valuable, saving valuable patents for possible offensive use against other DPL users. 199 Requiring all of a firm's patents to be offered under the DPL 199. Standard-setting organizations face similar gamesmanship problems. See Mark A. Lemley, Intellectual Property Rights and Standard-Setting Organizations, 90 CALIF. L. REV. 1889, 1912(2002 (describing strategic withdrawal from standard-setting organiza-

No. 1]
Protecting Open Innovation 57 guards against this. To limit the barrier to entry that the DPL all-in model creates, as well as to maximize freedom, DPL users may stop using the DPL at any time. However, again, in order to limit gamesmanship, they must give six months' notice before leaving, and all existing licenses remain in place. In other words, a DPL user may always cease offering future patents under the DPL, but must honor existing commitments. Further, during the six-month notice period, other DPL users may take any licenses needed from the leaving party, allowing them to plan for change and preventing the leaving party from packing up and stranding existing users. Finally, while the leaving party must honor any DPL commitment existing at the time it leaves, other DPL members may choose to revoke their licenses to it. Similarly, by "traveling with the patent" into any subsidiaries, affiliates and the like, the DPL reduces the ability of firms to use shell companies to engage in gamesmanship.
Perhaps the most complex set of reliability concerns relate to efforts at gamesmanship that involve the use of multiple entities to avoid the all-in requirement, the resale of DPL patents, or bankruptcy. 200 For example, entities motivated to attempt gamesmanship might try setting up shell corporations, transferring all of their offensive patents into those corporations, and then taking DPLs for their practicing entities, thereby gaining protection against OICs while allowing their shell corporations to sue them. While many versions of this scenario are possible with some difference in the details, all involve the use of multiple entities or patent transfers. For this reason, the DPL contains a broad successors-in-interest clause that allows DPL users to revoke any licenses of another DPL user if that user has economic or administrative control over a third party who files an offensive suit. While this cannot reach gamesmanship coordinated by truly independent actors, we believe that it is sufficient to prevent most forms of gamesmanship.
A related concern is that bankruptcy could upset the DPL equation by allowing a bankruptcy court to cancel DPLs on a patent that needed to be liquidated in order to satisfy creditors of the previous licensor. 201  Not only does the DPL have these safeguards in place, but it also allows more people to access the patents in the network because the price is zero for any license. Moreover, unless a pool has market power in a particular field, there is far less concern over anti-competitive outcomes. 206 Finally, a centralized model that relies on an ongoing entitywhich would need ongoing funding, staffing, and measures to ensure transparency to the community and to ensure that it is pursuing community goals -would raise separate reliability concerns. By using a FOSS-style, standardized cross-license as the organizing mechanism, the DPL model is intended to greatly lower administrative costs and eliminate the need for centralized decision-making, also limiting the need for ongoing trust in a centralized entity. While we envision a nonprofit running the centralized tracking website described above, there is no need for anyone to manage a costly centralized poollimiting the risk of resource constraints -and no need to rely on the judgment and values of one group or entity over time. Therefore, the DPL model is robustly defensive.

V. CONCLUSION
As OICs continue to increase their presence and influence across all areas of technology, the question of patent strategy will become increasingly important. It is our view that, in many cases, OICs cannot realistically avoid patents today. Instead, we believe that harnessing mechanisms, similar to those which OICs have used thus far to manage copyright risks, can help OICs employ defensive patenting to limit patent threats over time and eventually bring more stability to the patent system through private ordering. Therefore, in this Article, we