Chicago-Kent College of Law: Dean's Pages

UNBUNDLING VALUE IN ELECTRONIC INFORMATION PRODUCTS: INTELLECTUAL
PROPERTY PROTECTION FOR MACHINE READABLE INTERFACES
20 RUTGERS COMP. & TECH. L.J. 415

Henry H. Perritt, Jr. [FNa]

INTRODUCTION

Digital electronic network and optical storage technologies are revolutionizing the publishing industry. These electronic publishing [FN1] technologies facilitate publishing on demand, a process in which material from multiple sources is assembled and packaged in response to customized user requests. This process is fundamentally different from the traditional method of print publishing. Traditional print publishers begin with a package of information in raw form. They proceed to assemble the information, design its content and format, and manufacture its final form. All this effort is made in anticipation of and not in response to customer demand. The burden of evaluating and developing customer demand still falls on the traditional publishers' shoulders: they must market and promote their final product. Traditional technologies publish just in case while new technologies publish just in time.

*416 The new publishing technologies enable a disaggregation of value [FN2] and an associated dis-integration of production. [FN3] As a result, the different tasks of publishing will no longer have to be performed by the same party. A future supplier of raw content can be a different party from one who organizes the information, designs and sells retrieval systems, distributes the information, presents information, or markets the published material. [FN4] Additionally, other suppliers may be responsible for guaranteeing the quality of the separate aspects of value. This unbundling of value can lower barriers to entry for suppliers of value-added components of electronic publishing, and ensure continuing competition and innovation in the market.

The unbundling of different types of value contained in an information product presents serious intellectual property concerns. The body of traditional copyright law assumed that only authors and publishers were entitled to intellectual property protection. [FN5] Currently, the digital electronic information technologies make it possible for each type of value to be provided by a separate supplier assembled together in a customer requested format at the time the information is used. Thus, consumers are given the freedom to create compilations and derivative works on the fly as easily as any particular supplier.

It is not possible at this point to determine accurately how intellectual property and other legal incentives should be structured *417 to maximize the generation of value in new electronic information products. [FN6]

The new technologies add protection for investment as well as taking it away. They make possible some means of protecting markets that were never possible in a print on paper world, such as encryption and metering of usage, [FN7] but they make traditional intellectual property rights harder to enforce.

It is clear that the disaggregation of value reduces some of the tension that exists between protecting a first supplier's investment and encouraging the preparation of a wide variety of secondary works [FN8] that build on the first supplier's contribution. The new value in secondary works can be sold separately, allowing the consumer to combine the new value with the preexisting value. Disaggregation permits intellectual property incentives to be more narrowly targeted. However, for the benefits of disaggregation to be realized intellectual property rights must not impede the *418 compatibility of products. To reap the benefits of disaggregation the creation of a secondary work that is compatible with a preexisting work should not be considered an infringing act.

This Article begins by developing a taxonomy of ten different types of value in information products, which will facilitate the application of intellectual property concepts to the new publishing technologies. Next the paper develops the principal argument; that in order to promote independent production of associated types of value, legal protection of intellectual property rights must not extend to compatibility and machine-to-machine interfaces.

The Article's primary consideration is electronic information products which utilize public domain content, such as court opinions, statutes, and other information originating in government agencies. This approach helps to focus attention on incentives for producing alternative types of value, rather than focusing on producing content value. The production of alternative types of value is of particular importance in the electronic marketplace for information.

TYPES OF VALUE AND PRODUCT BUNDLES: LEVELS OF VALUE

Information products, whether produced and consumed in a paper or electronic form, are composed of ten different types of value. The value is bundled into information products through creating, organizing, retrieval-and-assembly, and marketing processes: [FN9]

-------------------------------------------------------------------------------

Process Type of Value Print Examples Electronic Examples

-------------------------------------------------------------------------------

Creating 1. content Content generated by same

[FN10] original author

-------------------------------------------------------------------------------

Organizing 2. chunking-an- Organizational same textual units,

d-tagging boundaries: tagged with typeface

sections, codes, footnote codes,

paragraphs, hypertext addresses;

pagination, chapter World Wide Web codes;

boundaries, headings WAIS headers

and titles, running

headers and footers,

page numbers

-------------------------------------------------------------------------------

Organizing 3. internal Table of contents, hypertext pointers to

pointers indexes other parts of same

information object;

World Wide Web

pointers

-------------------------------------------------------------------------------

Organizing 4. external Bibliographies hypertext pointers to

pointers other information

objects; menu

selections; file

names; Gopher pointers

-------------------------------------------------------------------------------

Retrieval & 5. presentation Print on paper video displays; WAIS

Assembly user interfaces

-------------------------------------------------------------------------------

Retrieval & 6. duplication All copies after the all copies after the

Assembly first first on diskette,

tape, or other media

-------------------------------------------------------------------------------

Retrieval & 7. distribution Getting the same

Assembly information from the

generator to the

consumer

-------------------------------------------------------------------------------

Marketing 8. promotion Advertising; inclusion same; inclusion in

in lists; product gateway lists and

reviews menus

-------------------------------------------------------------------------------

Marketing 9. billing Identifying users; same; metering usage of

assessing prices, particular files or

collecting money objects; automatic

confirmation of

billing authority like

credit cards

-------------------------------------------------------------------------------

Marketing 10. integrity-- Guaranteeing the same

assurance accuracy, expertise

of suppliers of each

type of value;

guaranteeing against

forgery or tampering

-------------------------------------------------------------------------------

*420 Suppliers and consumers of information printed on paper are familiar with each type of value. The value provided by chunking and tagging and internal and external pointers increases consumer utility by reducing the cost of human browsing, searching, and retrieving. Chunking and tagging value includes all basic typographic design features in print technologies. [FN11] For example, *421 browsing a newspaper is simple because the material has considerable chunking-and-tagging value reflected in headlines, separate stories, and the inverted pyramid style of journalistic writing. Newspapers with state-of-the art design also facilitate browsing through the use of indexes and internal pointer value. The internal pointer value in turn points to particular pages and story headlines, which constitute chunking-and- tagging value. Researchers utilize external pointer value when consulting the Index to the New York Times or the Readers' Guide to Periodical Literature. These sources contain external pointers to human-processable chunks and tags, call numbers, volume numbers, dates, page numbers, and titles of the articles. [FN12]

In electronic formats, chunking and tagging value works with the internal and external pointers value to permit the selection, coordination, and arrangement of subsets of content. A pointer points to a tag and retrieves the particular chunk with which the tag is associated. This is the functional equivalent of a conventional index pointing to a page number, the tag, and permitting the human user to retrieve the page, the chunk, to which the page number tag is associated. Pure chunking and tagging value is evidenced by a *422 blank form, an EDI transaction set, or any other data structure in a computer program.

Information products are bundles of these different value types. A book bundles value differently from a newspaper, which in turn is different from a law review article. A book bundles content value, which is reflected in the raw text, chunking-and-tagging value, which is reflected in its structure of articles, sections, pages, and paragraphs, internal pointers value in its structure of tables of contents and indexes, and presentation value in its method of printing all of this on paper bound between two covers. Each type of value adds utility to a consumer, principally by reducing the consumer's cost of using the information product. [FN13]

COMPETING BUNDLES OF VALUE

While new electronic information technologies facilitate the unbundling of the types of value, it is unlikely that each discrete type of value will be sold by itself. [FN14] Rather, the types of value *423 are likely to be sold in different and more numerous bundles than the bundles represented by a book. [FN15] The feasibility of bundling the different types of value in various ways is a potential enhancement to competition.

Electronic publishing, involving disaggregated production of different types of value, can be illustrated by the following hypothetical electronic publishing operation. The hypothetical builds logically on existing comprehensive commercial activities, [FN16] on the Supreme Court's Hermes initiative, [FN17] and on some distributed electronic publishing experiments underway at Cornell Law School, Villanova Law School, and at the Center for Computer Assisted Legal Instruction. The hypothetical enterprise might be called the Center for Distributed Electronic Publishing ("CenDEP"), which manages an FTP server connected to the Internet. [FN18]

CenDEP makes four different kinds of material available: (1) judicial opinions, (2) statutes, (3) student written law review articles and papers, and (4) practitioner-oriented textbooks and treatises. Some of these materials are already available electronically from several sources in word processing or ASCII form. All *424 of the materials as they now exist have, in addition to raw content, chunking and tagging value in the form of paragraph breaks, section numbers, and titles which indicate the start of major sections. This value was produced by the original author. [FN19] Their utility would be increased by adding more chunking and tagging, internal pointers, external pointers, and better presentation value.

CenDEP makes these materials retrievable according to cited statutory and constitutional sections, cited Restatement sections, the name of the justice authoring an opinion, and major legal concepts. CenDEP performs the search by employing additional chunking and tagging value in the following manner. CenDEP writes two computer programs. The first program, written in the Word for Windows macro language, defines four "styles," one for each of the following statutory citations, Restatement citations, Justice authoring an opinion segment, and substantive legal concepts. After the user enters the search criteria, the first program searches for character strings likely to appear in a statutory or constitutional citation, for example, "U.S.C." or "Stat.", and applies the first style to them. The second style is applied to every phrase that includes the word "Restatement." The third style is applied to names preceded by the words "Judge," "Justice," or "Circuit Judge," or followed by the letters "J.," or "C.J." The fourth style is applied to every instance of a word, in a set of words, expressing predefined legal concepts such as "independent contract," "Employment at Will," "discrimination," "tort," or "duty." [FN20] A person can take this program, which is a separate file and run it as a macro against any compatible file, including court opinions, statutes, articles and textbook material. The result of the process is a new file with additional chunking and tagging value.

*425 The second step is to process the Word for Windows files through Z39.50 (WAIS software). [FN21] This hypothetical WAIS software accepts textual documents in Word for Windows format and automatically imposes a structure on the text that permits its retrieval through free text queries and hypertext pointers, using a proprietary set of file formats and computer programs. [FN22] The WAIS structure follows the Word for Windows style tags added in the first step. The WAIS software also exports the information as a text file "marked-up" with SGML codes. [FN23]

CenDEP makes available, through Internet, both a WAIS interface and the SGML tagged files. CenDEP charges the user an hourly based fee for access, and sells client software to browse and retrieve the material.

The second hypothetical entrepreneur might be called "Another-DEP," or "ADEP" for short. ADEP makes a competing client front end, which uses the same underlying material, but only when an ADEP purchaser has a subscription with CenDEP. ADEP writes a computer program that presents the user with a series of menus to search for statutory citations, and to limit such a search to particular titles of the United States Code, Articles of the Constitution, Restatement citations, and labor and employment law *426 concepts. ADEP does not want to recreate the work already done by CenDEP. Moreover, CenDEP's programs are in wide use, and a substantial part of ADEP's expected market already uses CenDEP to rechunk and tag its court opinions, statutes, and secondary textual material obtained in word processing or ASCII formats from a variety of sources. Therefore, ADEP writes its program to search for the tags established by the CenDEP programs. This process requires ADEP to include a verbatim copy of the CenDEP Word for Windows style designation, the document type declaration and the SGML pointers in the ADEP program. The program described represents pure external pointers value, and may include some presentation value to enable a user to see highlighted search terms in the full text of a retrieved opinion.

The essential machine-to-machine interface question that the hypothetical presents is whether CenDEP infringes any copyright held by Microsoft, the vendor of Word for Windows, by searching for format codes, and whether ADEP's program infringes CenDEP's intellectual property rights in the Word for Windows style designations and style tags, or in the SGML document type declaration or SGML tags.

One could elaborate this model by supposing that subsequent entrepreneurs add promotion value and integrity assurance value. For example, a third entrepreneur called "OneStop" might establish an Internet server that would run both CenDEP programs and ADEP programs under a license. OneStop could offer a service that promotes the dialup availability of its server and adds integrity assurances, by spell and cite checking the material, before it is run through the programs and made available to the user.

The CenDEP/ADEP competition is just around the corner. Already, The Mead Corporation's LEXIS and West Publishing Corporation's WESTLAW make court opinions, most state and federal statutes, many law review articles and a few treatises available in centralized remotely accessible databases. Subscribers to the services have dialup access to these databases through packet switched digital networks provided by Mead and West or, at the user's option for a particular session, provided by Sprint, British Telecom, or CompuServe.

*427 The major difference between the CenDEP/ADEP hypothetical and the WESTLAW/LEXIS reality is the extent or degree of bundling. Both WESTLAW and LEXIS are completely bundled, except for content value. Both CenDEP and ADEP specialize in only certain types of value, and the interface between their complementary value types raises intellectual property questions.

Like modern print publishers, the early electronic publishers are brokers or assemblers of all of the types of value. A print publisher arranges with an author to supply content value, and hires designers and copy editors to supply chunking and tagging value. Substantial chunking and tagging value are also supplied by the author who typically designs the boundaries of chapter chunks, and determines the boundaries of section chunks. Publishers supervise extractors who prepare internal pointers value in the form of tables of contents and indexes, work in conjunction with the Library of Congress to arrange the cataloging-in-publication information and with Books in Print and reviewers to establish external pointers value. Publishers contract with printers and binders to supply duplication and presentation value, work with advertising agencies for promotion value, and handle distribution and billing value through warehousing and order fulfillment. Publishers traditionally also play an important quality control function, not only by editing and checking for clarity and accuracy, but also in the selection of material, that is, deciding which authors or concepts of information packaging are sufficiently authoritative to be useful to user communities.

New technologies may reduce the need for brokers of such broad scope. Advances in technology change the way types of value are bundled to meet user preferences and to reduce supplier costs. [FN24] Electronic technologies permit the unbundling of different aspects of added value, permitting each type, or clusters of only a few types, to be supplied separately to consumers and allow the consumer to assemble or bundle the different types of value on demand.

*428 Existing electronic publishers are discovering this new freedom. Historically, both WESTLAW and LEXIS sold totally bundled products. They had the raw information and supplied all the other values in the ten value type taxonomy. Recently, however, there has been a trend away from this practice. The electronic publishers now provide some gateway services, and there are new kinds of author relationships for both services. Cornell Law Professor Peter Martin's Electronic Social Security Treatise, for example, is available through LEXIS, and WESTLAW provides a new kind of gateway service for the Dialog database.

Over time, WESTLAW and LEXIS may concentrate on providing chunking and tagging and external pointers value, and rely more explicitly on public institutions to organize content in a rudimentary fashion and only provide value-added gateways to consumers. Modeled after the Martin relationship, they would offer different kinds of author relationships to obtain other kinds of content value. Potential authors would provide much smaller chunks of contributions that would be licensed - but not necessarily physically transferred - to WESTLAW or LEXIS, who would then add their own kind of value. The gradual acceptance of conventions for expressing chunk boundaries and tags, like SGML, facilitate unbundling by making it easier to achieve compatibility among different types of value. [FN25]

The attention given by the Clinton Administration to visions for a National Information Infrastructure is encouraging commercial enterprises to use Gopher, Archie, World Wide Web, and other Internet protocols that inherently disaggregate the production of information value, and use the electronic network as a market from which the consumer dictates the final assembly of a bundle of information value.

The point is not that each discrete type of value will be unbundled from all the others and sold separately. Rather, the points are, that each type of value should be thought about separately and that value types may be bundled differently from the *429 familiar print on paper bundles. In addition, disaggregation of value and a disintegration of production will result in greater competition and diversity among bundles, allowing greater consumer choice and greater reliance on the consumer to construct the ultimate bundle. Whether the law will allow a robust market in unbundled value to operate remains an unanswered question.

LAW AND ADDED VALUE

This Article considers a specific intellectual property issue: the predictability of computer readable interfaces in the context of disaggregated electronic publishing. Before examining the issue, it is necessary to establish a broader legal context.

All forms of intellectual property rights are justified by the need to protect creative or useful effort against free riding. [FN26] The risk of free riding depends on cost structures and the efficacy of technological protection. [FN27] The need for patent and copyright protection depends in part on the efficacy of contractual and trade secret protection. [FN28] The different forms of intellectual property rights protect different types of value in the overall bundle delivered to the consumer.

The need for legal protection of an innovator's investment depends on the risk of free riding. [FN29] The willingness to make a *430 significant investment in information value decreases as the potential for free riding increases. [FN30] When cost structures for producing a particular type of value encourage free riding and where the technological means of excluding free riders are impracticable, [FN31] legal protection is necessary if incentives are to *431 exist.

The law can protect markets in three basic ways: by enforcing private arrangements made between buyers and sellers, through contract law; by creating property interests in information value and allowing recovery of damages for, or injunctions against, invasions of those interests, through intellectual property law; and through tort law recovery for free riding accomplished by "wrongful" means. [FN32]

Different types of legal protection, which overlap but also maintain distinct foci, protect the different types of value. Copyright law tends to cover content, chunking and tagging and presentation value. [FN33] Trademark law tends to cover integrity assurance and promotion, [FN34] while patent law affords protection to the types of value falling in the middle, such as duplication and distribution. [FN35] Protection of internal and external pointers value, *432 and chunking and tagging and presentation processes is uncertain when they are disaggregated from content. These types of value are protected by trade secret, [FN36] as when the details of an interface [FN37] are kept secret, [FN38] or by contract, as in most remotely accessible database services.

Contract law, reinforced by trade secret concepts, can protect all kinds of value. However, contract protection has an important limitation. A contract protects only those who are parties to the contract. Absent another form of legal protection, a supplier of information value has no effective remedy against someone impairing his economic interest if he has no privity of contract with the actor. The supplier of content value may be able to enforce a contractual restriction on copying by his first customer, but if a customer of that customer copies, the original supplier has no remedy against the second downstream customer because there is no privity between that customer and the original supplier. Of *433 course, the original supplier may require his immediate customer to impose contractual restrictions on its customers, but a failure to impose such restrictions is simply a breach of contract by the first customer and creates no right against the subsequent customers not so restricted. Contract law thus creates limited protection against free riding. Only if a potential free rider must obtain something directly from the original producer is she likely to have privity of contract with the original producer. [FN39]

The Supreme Court's Feist decision [FN40] diminishes the scope of copyright protection for electronically published works. Before the Feist decision, the "sweat of the brow" doctrine [FN41] recognized in some judicial circuits [FN42] allowed copyright protection for types of value that may no longer be protectable. [FN43] "Sweat of the brow" protected the compilation effort of a compiler of a telephone directory by protecting against copying of the factual information contained in the telephone directory. [FN44] This was so even though *434 the facts themselves were not protected by copyright. [FN45] By implication, a compiler of an electronic database of judicial opinions and statutes could prohibit someone from copying the underlying public domain judicial opinions and statutes from his database even though the original information was not subject to copyright, and even though the amount of original authorship in the traditional sense involved in creating the database was de minimis. Conceptually, the "sweat of the brow" doctrine would protect suppliers against a free ride where the form of value enhanced utility to the consumer. Thus, a supplier who took raw content that was in the public domain and simply added duplication and distribution value would be entitled to recover for copyright infringement against a free rider who copied the public domain information at the point of delivery rather than adding his own duplication and distribution value.

Feist profoundly changes this approach in two ways. First, the preexisting information (for example, the names, addresses and telephone numbers in the case of the telephone book, or the judicial opinions and statutes in the case of WESTLAW or LEXIS) is no longer protectable merely because a compilation scheme or database structure has been superimposed on it. [FN46] Second, the new value added by the compiler is protectable only if some minimal amount of creative selection or arrangement is involved. [FN47] Copyright protection is no longer available simply to protect "sweat of the brow" value regardless of the eligibility of that kind of value for protection under the copyright statute. [FN48]

*435 After Feist, producers of duplication and distribution value are no longer entitled to copyright protection when they combine those increments of value with otherwise unprotectable expression. [FN49] Neither is chunking and tagging nor internal or external pointers protectable value unless the proponent of protection can show sufficient originality in the selection or arrangement, or both, represented by those types of value in a particular product. [FN50]

COMMON THEMES AND DILEMMAS

Patent and copyright protections are converging in high technology works. [FN51] Computer programs are utilitarian more than *436 expressive, [FN52] and the Supreme Court Feist opinion, the District Court opinion in Lotus and the Third Circuit opinion in Whelan all embrace a patent- like novelty threshold for the copyright originality elements roughly akin to, but less demanding than, the patent law novelty requirements. [FN53] Patent protection is being extended to compilations of data and to computer programs. [FN54] The utilitarian *437 character of computer produced and delivered works is more apparent when different types of value are unbundled.

Despite the convergence, the interests and subject matter protected by existing forms of intellectual property differ depending on the particular type of value. The Supreme Court's Feist opinion disavows intellectual property protection for entire bundles of information value, and targets intellectual property protection on the particular increments of value added by the person seeking protection. [FN55] The traditional subject matter of copyright is highly congruent with content, chunking and tagging, internal pointers, external pointers and presentation value. All of these are comfortably described as "expression." The subject matter of patent protection is most congruent with presentation, duplication, and billing value because patents can protect the processes that yield these types of value. According to the Uniform Trade Secrets Act ("UTSA") § 1(4), trade secret protection is applicable to the same instances as patent protection, and not beyond, because trade secret protection for other potential subject matter is lost when the subject matter is disclosed in the publishing process. [FN56]

*438 In recent years, [FN57] state courts have been willing to extend unfair competition and misappropriation concepts to protect information that otherwise is unprotected by traditional intellectual property law categories. [FN58] This trend draws upon International News Service v. Associated Press. [FN59] Building on this trend, Professor Wendy Gordon has proposed a comprehensive new tort that she calls "malcompetitive copying," that would afford broad legal protection for information, while minimizing harms resulting from undue restrictions on value adding producers who use protected information. [FN60] Gordon's scheme would achieve many of the same results as a scheme of broad protection accompanied by compulsory licensing, but it achieves it without the need for extensive statutory amendment. [FN61]

*439 AVOIDING BLOCKAGES TO NEW ADDITIONS OF VALUE

Historically, intellectual property protection reinforced technology, encouraging producers to bundle value. [FN62] As new information technologies make it easier to produce a single type of value without having to own and sell other types of value along with it, fewer inputs are required for a saleable product. The producer of one or two types of value need not pay to copy, adapt or distribute other types of value protected by intellectual property. [FN63] Thus technology now permits unbundling.

But unbundling permitted by the technology will occur only if intellectual property does not impede compatibility among different *440 types of value intended to be assembled by a consumer. New producers must be free to add external pointers, presentation and distribution value which draws upon preexisting value of other types. Producer B, wishing to construct a computerized topical guide to employment law, must be able to point to tags contained in a computer accessible employment law work belonging to A, without infringing on A's intellectual property. [FN64]

Affording broad intellectual property protection to low content works can foreclose access by other suppliers of value to basic data. [FN65] Similarly, affording intellectual property protection to interfaces diminishes the advantages to consumers of unbundling the type of information value now permitted by technology. Exclusive control over preparation of compatible works raises the cost to a consumer of buying one type of value and bundling it with other types of value contributed by other suppliers. It requires a supplier of subsequent value either to pay the other suppliers or to regenerate underlying value rather than simply using (without copying) what has been supplied by someone else.

Intellectual property law must permit a producer of one type of value to design his product to work in conjunction with another protected product containing primarily a different type of value. [FN66] The second producer's objective is compatibility, which is assured by having a common interface. [FN67]

*441 The first producer might argue that "appropriating" the interface infringes its intellectual property rights. Resolving this claim implicates two generic concerns: the scope of the "property" protected by the intellectual property doctrine, and the scope of the conduct with respect to that property that creates liability. [FN68] These concerns require consideration of the boundaries between protectable value and unprotectable ideas, the differences between facts and mental processes [FN69] and the variations of expression protectable by copyright and processes protectable by patent and trade secret. Recent decisions by the courts of appeals in Altai [FN70] and Sega, [FN71] and the directions of European Community law [FN72] correctly *442 leave interfaces between different types of value free for exploitation. But a recent decision by the district court deciding Lotus' claim against Borland allows an initial producer to monopolize the interface. [FN73]

PROTECTING INTERFACES - COMPATIBILITY AS INFRINGEMENT

Disaggregation potential and the associated potential for modular products make interfaces important. Through interfaces a subsequent producer builds on top of another producer's product without free riding on the value supplied by the first producer. [FN74]

The new world of disaggregated value will be realized only if intellectual property law does not discourage compatibility. [FN75] A *443 second supplier must be able to copy a previous supplier's data structures or query syntax so the second supplier's value can be matched up at the time of use with the first supplier's value; the two works must exchange information in compatible representations. [FN76] For example, unbundling can flourish only if the creator of external pointers value (or presentation or distribution value) is entitled to a limited kind of free ride on a separately existing and separately sold content and chunking and tagging value. This condition is a surrogate for the broader question of whether producers of separate types of value are entitled to free rides on preexisting value of other types as long as they do not copy and distribute the preexisting value, but only facilitate the use of such value by consumers. Some computer software copyright litigation raises doubts about whether copyright law permits this. In the hypo developed earlier in this article CenDEP would argue that ADEP, which directly produces only external pointers value, [FN77] has *444 infringed CenDEP's intellectual property. ADEP would not copy the content of the CenDEP material; that would be acquired by the user of the electronic book under the user's own subscription. But ADEP would copy parts of the CenDEP interface: the query syntax and the citations. Although CenDEP might have little economic incentive to object, because use of the electronic treatise would increase revenue to CenDEP theoretically, CenDEP could make three copyright arguments against ADEP: literal copying of the query language, [FN78] copying of protected structure, [FN79] and preparation of a derivative work. [FN80]

*445 The first argument is based on the proposition that the interface is a computer programming language or a computer program that is protected expression owned by the producer of the underlying work and that copying this interface is infringement. [FN81] This theory is best evaluated by considering the predictability of computer programming languages. [FN82] The second argument is based on the proposition that the interface represents protectable organization and structure of the underlying work. This argument is best evaluated by considering the reasoning of the Whelan, Altai and Lotus cases. The third argument is based on the proposition that *446 new value, like external pointers value, that has a compatible interface with preexisting value is a derivative work. This theory is best evaluated by considering the purpose of the derivative work protection. In addition to these three theories for extending intellectual property protection are three arguments against protection: public domain status of the interface, lack of free-riding involved, and fair use.

INTERMEDIATE COPYING

Production and marketing of disaggregated value in electronic works, like most of the recent video game interface cases, [FN83] is likely to involve some intermediate copying. In the video game cases it was necessary for the developer of a compatible work to copy certain parts of the original work in order to discover the interface. In almost every case of electronic publishing, some literal copying of tagging value in the chunking and tagging structure will be essential to implementation of a compatible interface. [FN84] If the interface, whether constituting a static data structure or a dynamic programming language, is protectable, then use of that interface involves literal copying. If such intermediate copying is an unprivileged copyright infringement, compatible works cannot be developed, produced and marketed without the *447 permission of the owner of the works embodying the types of value the new works are meant to complement.

The prevailing view in the video game cases is the correct one. Intermediate copying is a prima facie infringement as long as the material copied qualifies for protection. [FN85] The best way to permit the limited amount of intermediate copying necessary to produce a compatible work is through the fair use analysis. [FN86] Such privilege analysis makes it unnecessary to imply some new category of conduct - "intermediate copying" - not explicitly recognized in the statute.

COPYING OF PROTECTABLE LANGUAGE?

Some of the arguments made in Ashton Tate v. Fox Software [FN87] suggest that compatibility at the command and syntax level constitutes copyright infringement. [FN88]

*448 Ashton Tate's (now Borland's) dBASE IV is the market leader in desktop computer database management software. [FN89] Users of dBASE can interact with their data in one of two ways. They can make selections from menus contained in the software delivered by the vendor, or they can write programs in the dBASE programming language. The set of menus, the choices available, and the actions taken in response to user choices are protected, if at all, under the look and feel doctrine accepted in the Lotus case. [FN90] The Ashton Tate case involved a different set of questions relating to the predictability of the commands and syntax making up the dBASE programming language. [FN91]

The programming language works because Ashton Tate supplies an interpreter and a compiler, both of which are computer programs that take commands written in the dBASE programming language and translate them into lower level operations on the database, storage devices, output devices like video display screens and printers, and input devices like keyboards. The interpreter and compiler programs are protected expression. It would be an infringement for Fox or any other competitor to copy the programs and distribute them without Ashton Tate's permission. But that is not what Fox did. Fox wrote its own computer program, without copying Ashton Tate's programs, that accepts the same commands that are accepted by Ashton Tate's programs. The Fox software permits a user to write programs in the dBASE programming language just as though they were using dBASE software, but they buy Fox software instead. This availability of the Fox alternative *449 diminishes Ashton Tate's share of the market. Ashton Tate thus has an economic incentive to argue that the dBASE programming language is protectable, so it can exclude others from marketing systems that perform database operations according to dBASE commands. Lotus' position on the macro facility part of its case against Borland was similar to that of Ashton-Tate. When Borland made it possible for consumers with Lotus 123 macros to execute the macros under Quattro, it diminished the market for Lotus 123 that otherwise would be the only way to make the macros usable. [FN92] Fox's and Borland's position is remarkably like that of ADEP, which supplies hooks (external pointers) that fit into someone else's tags.

The programming language encompassing the command set is not, by itself, a computer program; rather, it is a set of acceptable commands and their syntax. Each command, when executed, triggers a process. A program written in the language, when executed, puts all the processes together into an application. Most of the commands are ordinary English words, the meaning of which bears a relationship to the action taken by the database management system when the commands are interpreted or compiled. For example, the word "find" is a command that finds database records meeting certain criteria. One cannot copyright the English language. It is in the public domain, and from a policy standpoint the restrictions on commerce resulting from intellectual property protection of ordinary words and phrases would be unacceptable. Furthermore, it is outside the scope of copyright because of the proviso in § 102(b) excluding ideas, concepts, processes and systems. [FN93]

Under Feist, however, selection and arrangement may be protectable even when the underlying raw material is not protectable. [FN94] Thus, Ashton Tate and Lotus argued that the command set and the syntax represent protectable selection and arrangement *450 from the larger set of English language words and English grammar syntax.

Independently, Ashton Tate argued that the command set and syntax were equivalent to a code or cipher, which is protectable regardless of whether the words in the code may have an ordinary meaning. As used in the code, or in the computer program command set, the words have a unique meaning and the association of the words in the command set with this unique meaning is an act involving substantial creativity. Ashton Tate sought not to prevent every copying, distribution, or derivative work of the word "find;" it sought only to prevent copying, distribution, or derivative works of the word "find" when it stands for this set of database operations developed and implemented by Ashton Tate in its dBASE language.

The prevailing view is that programming languages are notprotected by either intellectual property concept. [FN95] This consensus view has the advantage that it makes it easier for the producers of value-added electronic products to make those products compatible with underlying protected works and inventions.

COPYING PROTECTED STRUCTURE AND ORGANIZATION?

Apart from the literal interface copying arguments, a compatible interface may be an infringement if it copies other protectable expression. [FN96] If the structure and organization of the underlying value is protected by copyright, as Whelan Assoc. v. Jaslow Dental Lab. [FN97] and Atari [FN98] allow, and if designing a compatible interface *451 necessitates copying the structure, then compatibility constitutes infringement. [FN99]

The Third Circuit's Whelan case [FN100] stimulated a major controversy over the test for distinguishing unprotectable ideas from protectable expression. [FN101] Most of the post-Whelan cases have accepted the basic Learned Hand abstractions test for determining whether a merger between expression and idea has occurred so that the expression may not be protected without foreclosing use of the idea. [FN102] Most courts allow protection for the structure, sequence *452 and organization, and user interface, as long as there is room for different expressions of the function addressed by a particular structure, sequence, expression or user interface. [FN103]

In Computer Associates International, Inc. v. Altai, Inc., [FN104] the court of appeals held that certain interface similarities are unprotectable because they are dictated by operating system requirements. [FN105] Interfaces designed to be used by programmers also *453 may be unprotected because they are in the public domain. [FN106] The court found no infringement because all the similarities were either dictated by functional requirements or were in the public domain. [FN107] Of particular importance in the Altai decision is the principle that "compatibility requirements of other programs with which a program is designed to operate in conjunction" causes expression to be merged with idea and thus makes it unprotectable. [FN108]

Altai would exclude alternatives that are commercially inferior. [FN109] If the commercial objective is compatibility, there are no commercially acceptable alternatives except copying the interface. Indeed, the Altai court applied the merger conclusion to elements dictated by external factors, explicitly mentioning compatibility requirements of other programs and widely accepted programming practices within the industry. [FN110] Under this alternatives test, [FN111] an interface is not entitled to protection. There are no alternative ways to achieve compatibility between computerized works representing complementary types of value except by copying the interface. On efficiency grounds, viewed both from the perspective of the producer and the consumer, this single alternative means that *454 the interface should be treated as an idea rather than expression. [FN112]

There is one major problem with the Altai formulation that prevents it being a complete solution to the unbundled value interface problem. The compatibility efficiency/merger rationale of Altai does not necessarily extend to compatibility with the allegedly infringed work; the facts of Altai involved compatibility by both infringing and infringed programs with a third computing environment: the operating system. [FN113] On those facts, it was easier to conclude that the similar interfaces were dictated by external factors, given that the operating environments were external to both infringing and infringed programs. In the unbundled value scenario, the structure dictating the interface of the infringing program is not external; rather, it is the creative choice of the designer of the infringed program, not dictated by any desire of that original designer to be compatible with anything else.

The effort required to create an interface is significant and characterized by highly creative thought. It may be better analytically to consider the original designer's intended use for the interface than to have protection depend on whether he followed an industry standard or was creatively arbitrary.

In CMAX/Cleveland, Inc. v. UCR, Inc. [FN114] the significant and creative nature of the interface posed a problem for the defendant. A system of equipment rental stores developed an in-house store management program to replace one originally licensed from the plaintiff. [FN115] The plaintiff argued that the in-house program was either a copy of the plaintiff's program or a derivative work based on the plaintiff's program. [FN116] The defendant had copied screens and file structures in developing the in-house program. [FN117] The district court found that "the selection and arrangement of the field *455 definitions within the files ... are the expression of an idea." [FN118] The court rejected the defendant's argument that the file structures were dictated by market forces, concluding that the structures were not alphabetic or otherwise systematic nor were they functionally significant. [FN119] The court also found that transaction codes for use in the program were protectable expression rather than ideas because they were arbitrary and because they were not dictated by efficiency or by the industry. It rejected the defendant's argument that "it would not make sense to change the codes so the employees familiar with the earlier system would have to learn new ones." [FN120] Eliminating retraining costs was not an external factor sufficient to negate the copyrightability of the codes. [FN121]

The court, using the Computer Associates analytical framework, [FN122] including the external factors branch of the merger doctrine, [FN123] found infringement of protected expression based on evidence that the allegedly infringing file structures were not dictated by external factors. [FN124] CMAX thus used the Altai formula and reached a dif