Full opinion text
ORDER RE COPYRIGHTABILITY OF CERTAIN REPLICATED ELEMENTS OF THE JAVA APPLICATION PROGRAMMING INTERFACE WILLIAM ALSUP, District Judge. INTRODUCTION This action was the first of the so-called “smartphone war” cases tried to a jury. This order includes the findings of fact and conclusions of law on a central question tried simultaneously to the judge, namely the extent to which, if at all, certain replicated elements of the structure, sequence and organization of the Java application programming interface are protected by copyright. PROCEDURAL HISTORY In 2007, Google Inc., announced its Android software platform for mobile devices. In 2010, Oracle Corporation acquired Sun Microsystems, Inc., and thus acquired Sun’s interest in the popular programming language known as Java, a language used in Android. Sun was renamed Oracle America, Inc. Shortly thereafter, Oracle America (hereinafter simply “Oracle”) sued defendant Google and accused its Android platform as infringing Oracle’s Java-related copyrights and patents. Both Java and Android are complex platforms. Both include “virtual machines,” development and testing kits, and application programming interfaces, also known as APIs. Oracle’s copyright claim involves 37 packages in the Java API. Copyrightability of the elements replicated is the only issue addressed by this order. Due to complexity, the Court decided that the jury (and the judge) would best understand the issues if the trial was conducted in phases. The first phase covered copyrightability and copyright infringement as well as equitable defenses. The second phase covered patent infringement. The third phase would have dealt with damages but was obviated by stipulation and verdicts. For the first phase, it was agreed that the judge would decide issues of copyright-ability and Google’s equitable defenses and that the jury would decide infringement, fair use, and whether any copying was de minimis. Significantly, all agreed that Google had not literally copied the software but had instead come up with its own implementations of the 37 API packages. Oracle’s central claim, rather, was that Google had replicated the structure, sequence and organization of the overall code for the 37 API packages. For their task of determining infringement and fair use, the jury was told it should take for granted that the structure, sequence and organization of the 37 API packages as a whole was copyrightable. This, however, was not a final definitive legal ruling. One reason for this instruction was so that if the judge ultimately ruled, after hearing the phase one evidence, that the structure, sequence and organization in question was not protectable but was later reversed in this regard, the court of appeals might simply reinstate the jury verdict. In this way, the court of appeals would have a wider range of alternatives without having to worry about an expensive retrial. Counsel were so informed but not the jury. Each side was given seventeen hours of “air time” for phase one evidence (not counting openings, closings or motion practice). In phase one, as stated, the parties presented evidence on copyright-ability, infringement, fair use, and the equitable defenses. As to the compilable code for the 37 Java API packages, the jury found that Google infringed but deadlocked on the follow-on question of whether the use was protected by fair use. As to the documentation for the 37 Java API packages, the jury found no infringement. As to certain small snippets of code, the jury found only one was infringing, namely, the nine lines of code called “range-Check.” In phase two, the jury found no patent infringement across the board. (Those patents, it should be noted, had nothing to do with the subject addressed by this order.) The entire jury portion of the trial lasted six weeks. This order addresses and resolves the core premise of the main copyright claims, namely, whether the elements replicated by Google from the Java system were protectable by copyright in the first place. No law is directly on point. This order relies on general principles of copyright law announced by Congress, the Supreme Court and the Ninth Circuit. Counsel on both sides have supplied excellent briefing and the Court wishes to recognize their extraordinary effort and to thank counsel, including those behind the scenes burning midnight oil in law libraries, for their assistance. SUMMARY OF RULING So long as the specific code used to implement a method is different, anyone is free under the Copyright Act to write his or her own code to carry out exactly the same function or specification of any methods used in the Java API. It does not matter that the declaration or method header lines are identical. Under the rules of Java, they must be identical to declare a method specifying the same functionality' — even when the implementation is different. When there is only one way to express an idea or function, then everyone is free to do so and no one can monopolize that expression. And, while the Android method and class names could have been different from the names of their counterparts in Java and still have worked, copyright protection never extends to names or short phrases as a matter of law. It is true that the very same functionality could have been offered in Android without duplicating the exact command structure used in Java. This could have been done by re-arranging the various methods under different groupings among the various classes and packages (even if the same names had been used). In this sense, there were many ways to group the methods yet still duplicate the same range of functionality. But the names are more than just names — they are symbols in a command structure wherein the commands take the form java.package.Class.method() Each command calls into action a preassigned function. The overall name tree, of course, has creative elements but it is also a precise command structure' — a utilitarian and functional set of symbols, each to carry out a pre-assigned function. This command structure is a system or method of operation under Section 102(b) of the Copyright Act and, therefore, cannot be copyrighted. Duplication of the command structure is necessary for interoperability. STATEMENT OF FINDINGS 1. Java and Android. Java was developed by Sun, first released in 1996, and has become one of the world’s most popular programming languages and platforms. The Java platform, through the use of a virtual machine, enables software developers to write programs that are able to run on different types of computer hardware without having to rewrite them for each different type. Programs that run on the Java platform are written in the Java language. Java was developed to run on desktop computers and enterprise servers. The Java language, like C and C + +, is a human-readable language. Code written in a human-readable language — “source code” — is not readable by computer hardware. Only “object code,” which is not human-readable, can be used by computers. Most object code is in a binary language, meaning it consists entirely of Os and Is. Thus, a computer program has to be converted, that is, compiled, from source code into object code before it can run, or “execute.” In the Java system, source code is first converted into “byte-code,” an intermediate form, before it is then converted into binary machine code by the Java virtual machine. The Java language itself is composed of keywords and other symbols and a set of pre-written programs to carry out various commands, such as printing something on the screen or retrieving the cosine of an angle. The set of pre-written programs is called the application programming interface or simply API (also known as class libraries). In 2008, the Java API had 166 “packages,” broken into more than six hundred “classes,” all broken into over six thousand “methods.” This is very close to saying the Java API had 166 “folders” (packages), all including over six hundred pre-written programs (classes) to carry out a total of over six thousand subroutines (methods). Google replicated the exact names and exact functions of virtually all of these 37 packages but, as stated, took care to use different code to implement the six thousand-plus subroutines (methods) and six-hundred-plus classes. An API is like a library. Each package is like a bookshelf in the library. Each class is like a book on the shelf. Each method is like a how-to-do-it chapter in a book. Go to the right shelf, select the right book, and open it to the chapter that covers the work you need. As to the 37 packages, the Java and Android libraries are organized in the same basic way but all of the chapters in Android have been written with implementations different from Java but solving the same problems and providing the same functions. Every method and class is specified to carry out precise desired functions and, thus, the “declaration” (or “header”) line of code stating the specifications must be identical to carry out the given function. The accused product is Android, a software platform developed by Google for mobile devices. In August 2005, Google acquired Android, Inc., as part of a plan to develop a smartphone platform. Google decided to use the Java language for the Android platform. In late 2005, Google began discussing with Sun the possibility of taking a license to use and to adapt the entire Java platform for mobile devices. They also discussed a possible co-development partnership deal with Sun under which Java technology would become an open-source part of the Android platform, adapted for mobile devices. Google and Sun negotiated over several months, but they were unable to reach a deal. In light of its inability to reach agreement with Sun, Google decided to use the Java language to design its own virtual machine via its own software and to write its own implementations for the functions in the Java API that were key to mobile devices. Specifically, Google wrote or acquired its own source code to implement virtually all the functions of the 37 API packages in question. Significantly, all agree that these implementations — which account for 97 percent of the lines of code in the 37 API packages — are different from the Java implementations. In its final form, the Android platform also had its own virtual machine (the so-called Dalvik virtual machine), built with software code different from the code for the Java virtual machine. As to the 37 packages at issue, Google believed Java application programmers would want to find the same 37 sets of functionalities in the new Android system callable by the same names as used in Java.Code already written in the Java language would, to this extent, run on Android and thus achieve a degree of interoperability. The Android platform was released in 2007. The first Android phones went on sale the following year. Android-based mobile devices rapidly grew in popularity and now comprise a large share of the United States market. The Android platform is provided free of charge to smart-phone manufacturers. Google receives revenue through advertisement whenever a consumer uses particular functions on an Android smartphone. For its part, Sun and Oracle never successfully developed its own smartphone platform using Java technology. All agree that Google was and remains free to use the Java language itself. All agree that Google’s virtual machine is free of any copyright issues. All agree that the six-thousand-plus method implementations by Google are free of copyright issues. The copyright issue, rather, is whether Google was and remains free to replicate the names, organization of those names, and functionality of 37 out of 166 packages in the Java API, which has sometimes been referred to in this litigation as the “structure, sequence and organization” of the 37 packages. The Android platform has its own API. It has 168 packages, 37 of which are in contention. Comparing the 37 Java and Android packages side by side, only three percent of the lines of code are the same. The identical lines are those lines that specify the names, parameters and functionality of the methods and classes, lines called “declarations” or “headers.” In particular, the Android platform replicated the same package, method and class names, definitions and parameters of the 37 Java API packages from the Java 2SE 5.0 platform. This three percent is the heart of our main copyright issue. A side-by-side comparison of the 37 packages in the J2SE 5.0 version of Java versus in the Froyo version of Android shows that the former has a total of 677 classes (plus interfaces) and 6508 methods wherein the latter has 616 and 6088, respectively. Twenty-one of the packages have the same number of classes, interfaces and methods, although, as stated, the method implementations differ. The three percent of source code at issue includes “declarations.” Significantly, the rules of Java dictate the precise form of certain necessary lines of code called declarations, whose precise and necessary form explains why Android and Java must be identical when it comes to those particular lines of code. That is, since there is only one way to declare a given method functionality, everyone using that function must write that specific line of code in the same way. The same is true for the “calls,” the commands that invoke the methods. To see why this is so, this order will now review some of the key rules for Java programming. This explanation will start at the bottom and work its way upward. 2. The Java Language and Its API— Important Details. Java syntax includes separators {e.g., {, };), operators {e.g., /, <, >), literal values {e.g., 123, ‘x’, “Foo”), and keywords {e.g., if, else, while, return). These elements carry precise predefined meanings. Java syntax also includes identifiers {e.g., String, javadang.Object), which are used to name specific values, fields, methods, and classes as described below. These syntax elements are used to form statements, each statement being a single command executed by the Java compiler to take some action. Statements are run in the sequence written. Statements are commands that tell the computer to do work. A method is like a subroutine. Once declared, it can be invoked or “called on” elsewhere in the program. When a method is called on elsewhere in the program or in an application, “arguments” are usually passed to the method as inputs. The output from the method is known as the “return.” An example is a method that receives two numbers as inputs and returns the greater of the two as an output. Another example is a method that receives an angle expressed in degrees and returns the cosine of that angle. Methods can be much more complicated. A method, for example, could receive the month and day and return the Earth’s declination to the sun for that month and day. A method consists of the method header and the method body. A method header contains the name of the method; the number, order, type and name of the parameters used by the method; the type of value returned by the method; the checked exceptions that the method can throw; and various method modifiers that provide additional information about the method. At the trial, witnesses frequently referred to the method header as the “declaration.” This discrepancy has no impact on the ultimate analysis. The main point is that this header line of code introduces the method body and specifies very precisely its inputs, name and other functionality. Anyone who wishes to supply a method with the same functionality must write this line of code in the same way and must do so no matter how different the implementation may be from someone else’s implementation. The method body is a block of code that then implements the method. If a method is declared to have a return type, then the method body must have a statement and the statement must include the expression to be returned when that line of code is reached. During trial, many witnesses referred to the method body as the “implementation.” It is the method body that does the heavy lifting, namely the actual work of taking the inputs, crunching them, and returning an answer. The method body can be short or long. Google came up with its own implementations for the method bodies and this accounts for 97 percent of the code for the 37 packages. Once the method is written, tested and in place, it can be called on to do its work. A method call is a line of code somewhere else, such as in a different program that calls on (or invokes) the method and specifies the arguments to be passed to the method for crunching. The method would be called on using the command format “java.package.Class.method()” where () indicates the inputs passed to the method. For example, a = java.package.Class.method() would set the field “a” to equal the return of the method called. (The words “java.package.Class.method” would in a real program be other names like “java.lang.Math.max”; “java.package.Class.method” is used here simply to explain the format.) After a method, the next higher level of syntax is the class. A class usually includes fields that hold values (such as pi = 3.141592) and methods that operate on those values. Classes are a fundamental structural element in the Java language. A Java program is written as one or more classes. More than one method can be in a class and more than one class can be in a package. All code in a Java program must be placed in a class. A class declaration (or header) is a line that includes the name of the class and other information that define the class. The body of the class includes fields and methods, and other parameters. Classes can have subclasses that “inherit” the functionality of the class itself. When a new subclass is defined, the declaration line uses the word “extends” to alert the compiler that the fields and methods of the parent class are inherited automatically into the new subclass so that only additional fields or methods for the subclass need to be declared. The Java language does not allow a class to extend (be a subclass of) more than one parent class. This restrictiveness may be problematic when one class needs to inherit fields and methods from two different non-related classes. The Java programming language alleviates this dilemma through the use of “interfaces,” which refers to something different from the word “interface” in the API acronym. An interface is similar to a class. It can also contain methods. It is also in its own source code file. It can also be inherited by classes. The distinction is that a class may inherit from more than one interface whereas, as mentioned, a class can only inherit from one other class. For convenience, classes and interfaces are grouped into “packages” in the same way we all group files into folders on our computers. There is no inheritance function within packages; inheritance occurs only at the class and interface level. Here is a simple example of source code that illustrates methods, classes and packages. The italicized comments on the right are merely explanatory and are not compiled: package java.lang; public class Math { // Declares package javadang H Declares class Math H Declares method max I/ Implementation, returns x or I I Implementation, returns y H Closes method I I Closes class public static int max (int x, int y) { if (x > y) return x; else return y; } } To invoke this method from another program (or class), the following call could be included in the program: int a = java.lang.Math.max (2, 3); Upon reaching this statement, the computer would go and find the max method under the Math class in the java.lang package, input “2” and “3” as arguments, and then return a “3,” which would then be set as the value of “a.” The above example illustrates a point critical to our first main copyright issue, namely that the declaration line beginning “public static” is entirely dictated by the rules of the language. In order to declare a particular functionality, the language demands that the method declaration take a particular form. There is no choice in how to express it. To be specific, that line reads: public static int max (int x, int y) { The word “public” means that other programs can call on it. (If this instead says “private,” then it can only be accessed by other methods inside the same class.) The word “static” means that the method can be invoked without creating an instance of the class. (If this instead is an instance method, then it would always be invoked with respect to an object.) The word “int” means that an integer is returned by the method. (Other alternatives are “boolean,” “char,” and “String” which respectively mean “true/false,” “single character,” and “character string.”) Each of these three parameters is drawn from a short menu of possibilities, each possibility corresponding to a very specific functionality. The word “max” is a name and while any name (other than a reserved word) could have been used, names themselves cannot be copyrighted, as will be shown. The phrase “(int x, int y)” identifies the arguments that must be passed into the method, stating that they will be in integer form. The “x” and the “y” could be “a” and “b” or “argl” and “arg2,” so there is a degree of creativity in naming the arguments. Again, names cannot be copyrighted. (Android did not copy all of the particular argument names used in Java but did so as to some arguments.) Finally, is the beginning marker that tells the compiler that the method body is about to follow. The marker is mandatory. The foregoing description concerns the rules for the language itself. Again, each parameter choice other than the names has a precise functional choice. If someone wants to implement a particular function, the declaration specification can only be written in one way. Part of the declaration of a method can list any exceptions. When a program violates the semantic constraints of the Java language, the Java virtual machine will signal this error to the program as an exception for special handling. These are specified via “throw” statements appended at the end of a declaration. Android and Java are not identical in their throw designations but they are very similar as to the 37 packages at issue. A Java program must have at least one class. A typical program would have more than one method in a class. Packages are convenient folders to organize the classes. This brings us to the application programming interface. When Java was first introduced in 1996, the API included eight packages of pre-written programs. At least three of these packages were “core” packages, according to Sun, fundamental to being able to use the Java language at all. These packages were javadang, java, io, and java.util. As a practical matter, anyone free to use the language itself (as Oracle concedes all are), must also use the three core packages in order to make any worthwhile use of the language. Contrary to Oracle, there is no bright line between the language and the API. Each package was broken into classes and those in turn broken into methods. For example, javadang (a package) included Math (a class) which in turn included max (a method) to return the greater of two inputs, which was (and remains) callable as java.lang.Math.max with appropriate arguments (inputs) in the precise form required (see the example above). After Java’s introduction in 1996, Sun and the Java Community Process, a mechanism for developing a standard specifications for Java classes and methods, wrote hundreds more programs to carry out various nifty functions and they were organized into coherent packages by Sun to become the Java application programming interface. In 2008, as stated, the Java API had grown from the original eight to 166 packages with over six hundred classes with over six thousand methods. All of it was downloadable from Sun’s (now Oracle’s) website and usable by anyone, including Java application developers, upon agreement to certain license restrictions. Java was particularly useful for writing programs for use via the Internet and desktop computers. Although the declarations must be the same to achieve the same functionality, the names of the methods and the way in which the methods are grouped do not have to be the same. Put differently, many different API organizations could supply the same overall range of functionality. They would not, however, be interoperable. Specifically, code written for one API would not run on an API organized differently, for the name structure itself dictates the precise form of command to call up any given method. To write a fresh program, a programmer names a new class and adds fields and methods. These methods can call upon the pre-written functions in the API. Instead of re-inventing the wheels in the API from scratch, programmers can call on the tried-and-true pre-packaged programs in the API. These are ready-made to perform a vast menu of functions. This is the whole point of the API. For example, a student in high school can write a program that can call upon java.lang.Math.max to return the greater of two numbers, or to find the cosine of an angle, as one step in a larger homework assignment. Users and developers can supplement the API with their own specialized methods and classes. The foregoing completes the facts necessary to decide the copyrightability issue but since Oracle has made much of two small items copied by Google, this order will now make findings thereon so that there will be proper context for the court of appeals. 3. Rangecheck and the De-Compiled Test Files. Oracle has made much of nine lines of code that crept into both Android and Java. This circumstance is so innocuous and overblown by Oracle that the actual facts, as found herein by the judge, will be set forth below for the benefit of the court of appeals. Dr. Joshua Bloch worked at Sun from August 1996 through July 2004, eventually holding the title of distinguished engineer. While working at Sun, Dr. Bloch wrote a nine-line code for a function called “range-Check,” which was put into a larger file, “Arrays.java,” which was part of the class library for the 37 API packages at issue. The function of rangeCheck was to check the range of a list of values before sorting the list. This was a very simple function. In 2004, Dr. Bloch left Sun to work at Google, where he came to be the “chief Java architect” and “Java guru.” Around 2007, Dr. Bloch wrote the files, “Tim-sort.java” and “ComparableTimsort,” both of which included the same rangeCheck function he wrote while at Sun. He wrote the Timsort files in his own spare time and not as part of any Google project. He planned to contribute Timsort and ComparableTimsort back to the Java community by submitting his code to an open implementation of the Java platform, OpenJDK, which was controlled by Sun. Dr. Bloch did, in fact, contribute his Timsort file to OpenJDK and Sun included Timsort as part of its Java J2SE 5.0 release. In 2009, Dr. Bloch worked on Google’s Android project for approximately one year. While working on the Android team, Dr. Bloch also contributed Timsort and ComparableTimsort to the Android platform. Thus, the nine-line rangeCheck function was copied into Google’s Android. This was how the infringement happened to occur. When discovered, the range-Check lines were taken out of the then-current version of Android over a year ago. The rangeCheck block of code appeared in a class containing 3,179 lines of code. This was an innocent and inconsequential instance of copying in the context of a massive number of lines of code. Since the remainder of this order addresses only the issue concerning structure, sequence and organization, and since rangeCheck has nothing to do with that issue, rangeCheck will not be mentioned again, but the reader will please remember that it has been readily conceded that these nine lines of code found their way into an early version of Android. Google also copied eight computer files by decompiling the bytecode from eight Java files back into source code and then using the source code. These files were merely used as test files and never found their way into Android or any handset. These eight files have been treated at trial as a single unit. Line by line, Oracle tested all fifteen million lines of code in Android (and all files used to test along the way leading up to the final Android) and these minor items were the only items copied, save and except for the declarations and calls which, as stated, can only be written in one way to achieve the specified functionality. ANALYSIS AND CONCLUSIONS OF LAW 1. Names and Short Phrases. To start with a clear-cut rule, names, titles and short phrases are not copyrightable, according to the United States Copyright Office, whose rule thereon states as follows: Copyright law does not protect names, titles, or short phrases or expressions. Even if a name, title, or short phrase is novel or distinctive or lends itself to a play on words, it cannot be protected by copyright. The Copyright Office cannot register claims to exclusive rights in brief combinations of words such as: • Names of products or services. • Names of business organizations, or groups (including the names of performing groups). • Pseudonyms of individuals (including pen or stage names). • Titles of works. • Catchwords, catchphrases, mottoes, slogans, or short advertising expressions. • Listings of ingredients, as in recipes, labels, or formulas. When a recipe or formula is accompanied by an explanation or directions, the text directions may be copyrightable, but the recipe or formula itself remains uncopyrightable. U.S. Copyright Office, Circular 34; see 37 C.F.R. 202.1(a). This rule is followed in the Ninth Circuit. Sega Enters., Ltd. v. Accolade, Inc., 977 F.2d 1510, 1524 n. 7 (9th Cir.1992). This has relevance to Oracle’s claim of copyright ownership over names of methods, classes and packages. 2. The Development of Law on the Copyrightability of Computer Programs and Their Structure, Sequence and Organization. Turning now to the more difficult question, this trial showcases a distinction between copyright protection and patent protection. It is an important distinction, for copyright exclusivity lasts 95 years whereas patent exclusivity lasts twenty years. And, the Patent and Trademark Office examines applications for anticipation and obviousness before allowance whereas the Copyright Office does not. This distinction looms large where, as here, the vast majority of the code was not copied and the copyright owner must resort to alleging that the accused stole the “structure, sequence and organization” of the work. This phrase — structure, sequence and organization — does not appear in the Act or its legislative history. It is a phrase that crept into use to describe a residual property right where literal copying was absent. A question then arises whether the copyright holder is more appropriately asserting an exclusive right to a functional system, process, or method of operation that belongs in the realm of patents, not copyrights. A. Baker v. Seldom. The general question predates computers. In the Supreme Court’s decision in Baker v. Selden, 101 U.S. 99, 25 L.Ed. 841 (1879), the work at issue was a book on a new system of double-entry bookkeeping. It included blank forms, consisting of ruled lines, and headings, illustrating the system. The accused infringer copied the method of bookkeeping but used different forms. The Supreme Court framed the issue as follows: The evidence of the complainant is principally directed to the object of showing that Baker uses the same system as that which is explained and illustrated in Selden’s books. It becomes important, therefore, to determine whether, in obtaining the copyright of his books, he secured the exclusive right to the use of the system or method of book-keeping which the said books are intended to illustrate and explain. Id. at 101. Baker held that using the same accounting system would not constitute copyright infringement. The Supreme Court explained that only patent law can give an exclusive right to a method: To give to the author of the book an exclusive property in the art described therein, when no examination of its novelty has ever been officially made, would be a surprise and a fraud upon the public. That is the province of letters-patent, not of copyright. The claim to an invention or discovery of an art or manufacture must be subjected to the examination of the Patent Office before an exclusive right therein can be obtained; and it can only be secured by a patent from the government. Id. at 102. The Supreme Court went on to explain that protecting the method under copyright law would frustrate the very purpose of publication: The copyright of a work on mathematical science cannot give to the author an exclusive right to the methods of operation which he propounds, or to the diagrams which he employs to explain them, so as to prevent an engineer from using them whenever occasion requires. The very object of publishing a book on science or the useful arts is to communicate to the world the useful knowledge which it contains. But this object would be frustrated if the knowledge could not be used without incurring the guilt of piracy of the book. Id. at 103. Baker also established the “merger” doctrine for systems and methods intermingled with the texts or diagrams illustrating them: And where the art it teaches cannot be used without employing the methods and diagrams used to illustrate the book, or such as are similar to them, such methods and diagrams are to be considered as necessary incidents to the art, and given therewith to the public; not given for the purpose of publication in other works explanatory of the art, but for the purpose of practical application. Ibid. It is true that Baker is aged but it is not passé. To the contrary, even in our modern era, Baker continues to be followed in the appellate courts, as will be seen below. B. The Computer Age and Section 102(b) of the 1976 Act. Almost a century later, Congress revamped the Copyright Act in 1976. By then, software for computers was just emerging as a copyright issue. Congress decided in the 1976 Act that computer programs would be copyrightable as “literary works.” See H.R. Rep. No. 94-1476, at 54 (1976), 1976 U.S.C.C.A.N. 5659, 5667. There was, however, no express definition of a computer program until an amendment in 1980. The 1976 Act also codified a Baker-like limitation on the scope of copyright protection in Section 102(b). See Apple Computer, Inc. v. Microsoft Corp., 35 F.3d 1435, 1443 n. 11 (9th Cir.1994). Section 102(b) stated (and still states): In no case does copyright protection for an original work of authorship extend to any idea, procedure, process, system, method of operation, concept, principle, or discovery, regardless of the form in which it is described, explained, illustrated, or embodied in such work. The House Report that accompanied Section 102(b) of the Copyright Act explained: Copyright does not preclude others from using the ideas or information revealed by the author’s work. It pertains to the literary, musical, graphic, or artistic form in which the author expressed intellectual concepts. Section 102(b) makes clear that copyright protection does not extend to any idea, procedure, process, system, method of operation, concept, principle, or discovery, regardless of the form in which it is described, explained, illustrated, or embodied in such work. Some concern has been expressed lest copyright in computer programs should extend protection to the methodology or processes adopted by the programmer, rather than merely to the ‘writing’ expressing his ideas. Section 102(b) is intended, among other things, to make clear that the expression adopted by the programmer is the copyrightable element in a computer program, and that the actual processes or methods embodied in the program are not within the scope of the copyright law. Section 102(b) in no way enlarges or contracts the scope of copyright protection under the present law. Its purpose is to restate, in the context of the new single Federal system of copyright, that the basic dichotomy between expression and idea remains unchanged. H.R. Rep. No. 94-1476, at 56-57 (1976), 1976 U.S.C.C.A.N. 5659, 5670 (emphasis added). Recognizing that computer programs posed novel copyright issues, Congress established the National Commission on New Technological Uses of Copyrighted Works (referred to as CONTU) to recommend the extent of copyright protection for software. The Commission consisted of twelve members with Judge Stanley Fuld as chairman and Professor Melville Nimmer as vice-chairman. The Commission recommended that a definition of “computer program” be added to the copyright statutes. This definition was adopted in 1980 and remains in the current statute: A “computer program” is a set of statements or instructions to be used directly or indirectly in a computer in order to bring about a certain result. 17 U.S.C. § 101. Moreover, the CON-TU report stated that Section 102(b)’s preclusion of copyright protection for “procedure, process, system, method of operation” was reconcilable with the new definition of “computer program.” The Commission explained the dichotomy between copyrightability and non-eopyrightability as follows: Copyright, therefore, protects the program so long as it remains fixed in a tangible medium of expression but does not protect the electromechanical functioning of a machine. The way copyright affects games and game-playing is closely analogous: one may not adopt and republish or redistribute copyrighted game rules, but the copyright owner has no power to prevent others from playing the game. Thus, one is always free to make a machine perform any conceivable process (in the absence of a patent), but one is not free to take another’s program. Nat’l Comm’n on New Technological Uses of Copyrighted Works, Final Report 20 (1979) (emphasis added). The Commission also recognized the “merger” doctrine, a rule of importance a few pages below in this order (emphasis added): The “idea-expression identity” exception provides that copyrighted language may be copied without infringing when there is but a limited number of ways to express a given idea. This rule is the logical extension of the fundamental principle that copyright cannot protect ideas. In the computer context this means that when specific instructions, even though previously copyrighted, are the only and essential means of accomplishing a given task, their later use by another will not amount to an infringement .... [C]opyright protection for programs does not threaten to block the use of ideas or program language previously developed by others when that use is necessary to achieve a certain result. When other language is available, programmers are free to read copyrighted programs and use the ideas embodied in them in preparing their own works. Ibid. The Commission realized that differentiating between the copyrightable form of a program and the uncopyrightable process was difficult, and expressly decided to leave the line drawing to federal courts: [T]he many ways in which programs are now used and the new applications which advancing technology will supply may make drawing the line of demarcation more and more difficult. To attempt to establish such a line in this report written in 1978 would be futile .... Should a line need to be drawn to exclude certain manifestations of programs from copyright, that line should be drawn on a case-by-case basis by the institution designed to make fine distinctions — the federal judiciary. Id. at 22-23. Congress prepared no legislative reports discussing the CONTU comments regarding Section 102(b). See H.R. Rep. No. 96-1307, at 23-24 (1980). Nevertheless, Congress followed CONTU’s recommendations by adding the definition of computer programs to the statute and amending a section of the Act not relevant to this order. See Apple Computer, Inc. v. Formula Intern. Inc., 725 F.2d 521, 522-25 (9th Cir. 1984). Everyone agrees that no one can copy line-for-line someone else’s copyrighted computer program. When the line-by-line listings are different, however, some copyright owners have nonetheless accused others of stealing the “structure, sequence and organization” of the copyrighted work. That is the claim here. C. Decisions Outside the Ninth Circuit. No court of appeals has addressed the copyrightability of APIs, much less their structure, sequence and organization. Nor has any district court. Nevertheless,, a review of the case law regarding non-literal copying of software provides guidance. Circuit decisions outside the Ninth Circuit will be considered first. The Third Circuit led off in Whelan Associates, Inc. v. Jaslow Dental Laboratory, Inc., 797 F.2d 1222 (3d Cir.1986). In that case, the claimant owned a program, Dentalab, that handled the administrative and bookkeeping tasks of dental prosthetics businesses. The accused infringer developed another program, Dentcom, using a different programming language. The Dentcom program handled the same tasks as the Dentalab program and had the following similarities: The programs were similar in three significant respects ... most of the file structures, and the screen outputs, of the programs were virtually identical ... five particularly important “subroutines” within both programs — order entry, invoicing, accounts receivable, end of day procedure, and end of month procedure — performed almost identically in both programs. Id. at 1228. On these facts, the district court had found, after a bench trial, that the accused infringer copied the claimant’s software program. Id. at 1228-29. On appeal, the accused infringer argued that the structure of the claimant’s program was not protectable under copyright. In rejecting this argument, the court of appeals created the following framework to deal with non-literal copying of software: [T]he line between idea and expression may be drawn with reference to the end sought to be achieved by the work in question. In other words, the purpose or function of a utilitarian work would be the work’s idea, and everything that is not necessary to that purpose or function would be part of the expression of the idea. Id. at 1236 (emphasis in original). Applying this test, Whelan found that the structure of Dentalab was copyrightable because there were many different ways to structure a program that managed a dental laboratory: [T]he idea of the Dentalab program was the efficient management of a dental laboratory (which presumably has significantly different requirements from those of other businesses). Because that idea could be accomplished in a number of different ways with a number of different structures, the structure of the Dentalab program is part of the program’s expression, not its idea. Id. at 1236 n. 28. The phrase “structure, sequence and organization” originated in a passage in Whelan explaining that the opinion used those words interchangeably and that, although not themselves part of the Act, they were intended to capture the thought that “sequence and order could be parts of the expression, not the idea, of a work.” Id. at 1239,1248. To summarize, in affirming the district court’s final judgment of infringement, Whelan held that the structure of the Dentalab program was copyrightable because there were many other ways to perform the same function of handling the administrative and bookkeeping tasks of dental prosthetics businesses with different structures and designs. Id. at 1238. Others were free to come up with their own version but could not appropriate the Dentalab structure. This decision plainly seems to have been the high-water mark of copyright protection for the structure, sequence and organization of computer programs. It was also the only appellate decision found by the undersigned judge that affirmed (or directed) a final judgment of copyrightability on a structure, sequence and organization theory. Perhaps because it was the first appellate decision to wade into this problem, Whelan has since been criticized by subsequent treatises, articles, and courts, including our own court of appeals. See Sega Enters., Ltd. v. Accolade, Inc., 977 F.2d 1510, 1524-25 (9th Cir.1992). Instead, most circuits, including ours, have adopted some variation of an approach taken later by the Second Circuit. See Apple Computer, Inc. v. Microsoft Corp., 35 F.3d 1435, 1445 (9th Cir.1994). In Computer Associates International, Inc. v. Altai 982 F.2d 693 (2d Cir.1992), the claimant owned a program designed to translate the language of another program into the particular language that the computer’s operating system would be able to understand. The accused infringer developed its own program with substantially similar structure but different source code (using the same programming language). The Second Circuit criticized Whelan for taking too narrow a view of the “idea” of a program. The Second Circuit adopted instead an “abstract-filtration-comparison” test. The test first dissected the copyrighted program into its structural components: In ascertaining substantial similarity under [the abstract-filtration-comparison test], a court would first break down the allegedly infringed program into its constituent structural parts. Then, by examining each of these parts for such things as incorporated ideas, expression that is necessarily incidental to those ideas, and elements that are taken from the public domain, a court would then be able to sift out all non-protectable material. Id. at 706. Then, the test filtered out structures that were not copyrightable. For this filtration step, the court of appeals relied on the premise that programmers fashioned structures “to maximize the program’s speed, efficiency, as well as simplicity for user operation, while taking into consideration certain externalities such as the memory constraints of the computer upon which the program will be run.” Id. at 698. Because these were “practical considerations,” the court held that structures based on these considerations were not copyrightable expressions. Thus, for the filtration step, the court of appeals outlined three types of structures that should be precluded from copyright protection. First, copyright protection did not extend to structures dictated by efficiency. A court must inquire whether the use of this particular set of modules [is] necessary efficiently to implement that part of the program’s process being implemented. If the answer is yes, then the expression represented by the programmer’s choice of a specific module or group of modules has merged with their underlying idea and is unprotected. Id. at 708 (emphasis in original). Paradoxically, this meant that non-efficient structures might be copyrightable while efficient structures may not be. Nevertheless, the Second Circuit explained its reasoning as follows: In the context of computer program design, the concept of efficiency is akin to deriving the most concise logical proof or formulating the most succinct mathematical computation. Thus, the more efficient a set of modules are, the more closely they approximate the idea or process embodied in that particular aspect of the program’s structure. While, hypothetically, there might be a myriad of ways in which a programmer may effectuate certain functions within a program — i.e., express the idea embodied in a given subroutine — efficiency concerns may so narrow the practical range of choice as to make only one or two forms of expression workable options. Ibid. Efficiency also encompassed user simplicity and ease of use. Id. at 708-09. Second, copyright protection did not extend to structures dictated by external factors. The court explained this as follows: [I]n many instances it is virtually impossible to write a program to perform particular functions in a specific computing environment without employing standard techniques. This is a result of the fact that a programmer’s freedom of design choice is often circumscribed by extrinsic considerations such as (1) the mechanical specifications of the computer on which a particular program is intended to run; (2) compatibility requirements of other programs with which a program is designed to operate in conjunction; (3) computer manufacturers’ design standards; (4) demands of the industry being serviced; and (5) widely accepted programming practices within the computer industry. Id. at 709-10. Third, copyright protection did not extend to structures already found in the public domain. The court reasoned that materials in the public domain, such as elements of a computer program that have been freely accessible, cannot be appropriated. Ibid. Ultimately, in the case before it, the Second Circuit held that after removing unprotectable elements using the criteria discussed above, only a few lists and macros in accused product were similar to the copied product, and their impact on the program was not large enough to declare copyright infringement. Id. at 714-15. The copyright claim, in short, failed. The Tenth Circuit elaborated on the abstract-filtration-comparison test in Gates Rubber Co. v. Bando Chemical Industries, Ltd., 9 F.3d 823 (10th Cir.1993). There, the claimant developed a computer program that determined the proper rubber belt for a particular machine by performing complicated calculations involving numerous variables. The program used published formulas in conjunction with certain mathematical constants developed by the claimant to determine belt size. The Tenth Circuit offered the following description of a software program’s structure: The program’s architecture or structure is a description of how the program operates in terms of its various functions, which are performed by discrete modules, and how each of these modules interact with each other. Id. at 835. As had the Second Circuit, the Tenth Circuit held that filtration should eliminate the unprotectable elements of processes, facts, public domain information, merger material, scenes a faire material, and other unprotectable elements suggested by the particular facts of the program under examination. For Section 102(b) processes, the court gave the following description: Returning then to our levels of abstraction framework, we note that processes can be found at any level, except perhaps the main purpose level of abstraction. Most commonly, processes will be found as part of the system architecture, as operations within modules, or as algorithms. Id. at 837. The court described the scenes a faire doctrine for computer programs as follows: The scenes a faire doctrine also excludes from protection those elements of a program that have been dictated by external factors. In the area of computer programs these external factors may include: hardware standards and mechanical specifications, software standards and compatibility requirements, Sega Enterprises Ltd. v. Accolade, Inc., 977 F.2d 1510, 1525-27 (9th Cir.1993), computer manufacturer design standards, target industry practices and demands, and computer industry programming practices. We recognize that the scenes a faire doctrine may implicate the protectability of interfacing and that this topic is very sensitive and has the potential to effect [sic] widely the law of computer copyright. This appeal does not require us to determine the scope of the scenes a faire doctrine as it relates to interfacing and accordingly we refrain from discussing the issue. Id. at 838 & n. 14 (all citations omitted except Sega). Like the Second Circuit, the Tenth Circuit also listed many external considerations — such as compatibility, computer industry programming practices, and target industry practices and demands — that would exclude elements from copyright protection under the scenes a faire doctrine. Ultimately, the Tenth Circuit remanded because the district court had failed to make specific findings that fit this framework. The First Circuit weighed in with its 1995 decision Lotus Development Corp. v. Borland International, Inc., 49 F.3d 807 (1st Cir.1995). In Lotus, the claimant owned the Lotus 1-2-3 spreadsheet program that enabled users to perform accounting functions electronically on a computer. Users manipulated and controlled the program via a series of menu commands, such as “Copy,” “Print,” and “Quit.” In all, Lotus 1-2-3 had 469 commands arranged into more than 50 menus and submenus. Lotus 1-2-3 also allowed users to write “macros,” whereby a user could designate a series of command choices (sequence of menus and submenus) with a single macro keystroke. Then, to execute that series of commands, the user only needed to type the single pre-programmed macro keystroke, causing the program to recall and perform the designated series of commands automatically. Id. at 809-10. The accused infringer Borland developed a competing spreadsheet program. Borland included the Lotus menu command hierarchy in its program to make it compatible with Lotus 1-2-3 so that spreadsheet users who were already familiar with Lotus 1-2-3 would be able to switch to the Borland program without having to learn new commands or rewrite their Lotus macros. In so doing, Borland did not copy any of Lotus’s underlying source or object code. (The opinion did not say whether the programs were written in the same language.) The district court had ruled that the Lotus 1-2-3 menu command hierarchy was a copyrightable expression because there were many ways to construct a spreadsheet menu tree. Thus, the district court had concluded that the Lotus developers’ choice and arrangement of command terms, reflected in the Lotus menu command hierarchy, constituted copyrightable expression. Id. at 810-11. The First Circuit, however, held that the Lotus menu command hierarchy was not copyrightable because it was a method of operation under Section 102(b). The court explained: We think that “method of operation,” as that term is used in § 102(b), refers to the means by which a person operates something, whether it be a car, a food processor, or a computer. Thus a text describing how to operate something would not extend copyright protection to the method of operation itself; other people would be free to employ that method and to describe it in their own words. Similarly, if a new method of operation is used rather than described, other people would still be free to employ or describe that method. Id. at 815. The court reasoned that because the menu command hierarchy was essential to make use of the program’s functional capabilities, it should be properly categorized as a “method of operation” under Section 102(b). The court explained: The Lotus menu command hierarchy does not merely explain and present Lotus 1-2-3’s functional capabilities to the user; it also serves as the method by which the program is operated and controlled ----In other words, to offer the same capabilities as Lotus 1-2-3, Borland did not have to copy Lotus’s underlying code (and indeed it did not); to allow users to operate its programs in substantially the same way, however, Borland had to copy the Lotus menu command hierarchy. Thus the Lotus 1-2-3 code is not a uneopyrightable “method of operation.” Ibid. Thus, the court reasoned that although Lotus had made “expressive” choices of what to name the command terms and how to structure their hierarchy, it was nevertheless an uncopyrightable “method of operation.” The Lotus decision was affirmed by an evenly divided Supreme Court (four to four). The Federal Circuit had the opportunity to apply Lotus in an appeal originating from the District of Massachusetts in Hutchins v. Zoll Medical Corp., 492 F.3d 1377 (Fed.Cir.2007) (affirming summary judgment against copyright owner). In Hutchins, the claimant owned a program for performing CPR and argued that his copyright covered the “system of logic whereby CPR instructions are provided by computerized display, and [] the unique logic contained in [his] software program.” Id. at 1384. The claimant argued that the accused program was similar because it “performfed] the same task in the same way, that is, by measuring heart activity and signaling the quantity and timing of CPR compressions to be performed by the rescuer.” Ibid. The court of appeals rejected this argument, holding that copyright did not protect the “technologic method of treating victims by using CPR and instructing how to use CPR.” Ibid. (citing Lotus). D. Decisions in the Supreme Court and in our Circuit. Our case is governed by the law in the Ninth Circuit and, of course, the Supreme Court. The Supreme Court missed the opportunity to address these issues in Lotus due to the four-to-four affirmance and has, thus, never reached the general question. Nonetheless, Baker, which is still good law, provides guidance and informs how we should read Section 102(b). Another Supreme Court decision, Feist Publications, Inc. v. Rural Telephone Service Co., Inc., 499 U.S. 340, 111 S.Ct. 1282, 113 L.Ed.2d 358 (1991), which dealt primarily with the eopyrightability of purely factual compilations, provided some general principles. In Feist, the Supreme Court considered the eopyrightability of a telephone directory comprised of names, addresses, and phone numbers organized in alphabetical order. The Supreme Court rejected the notion that copyright law was meant to reward authors for the “sweat of the brow.” This meant that we should not yield to the temptation to award copyright protection merely because a lot of sweat went into the work. The Supreme Court concluded that protection only extended to the original components of an author’s work. Id. at 353, 111 S.Ct. 1282. The Supreme Court concluded: This inevitably means that the copyright in a factual compilation is thin. Notwithstanding a valid copyright, a subsequent compiler remains free to use the facts contained in another’s publication to aid in preparing a competing work, so long as the competing work does not feature the same selection and arrangement. Id. at 349, 111 S.Ct. 1282. Turning to our own Ninth Circuit, our court of appeals has recognized that non-literal components of a program, including the structure, sequence and organization and user interface, can be protectable under copyright depending on whether the structure, sequence and organization in question qualifies as an expression of an idea rather than an idea itself. Johnson Controls, Inc. v. Phoenix Control Sys., Inc., 886 F.2d 1173, 1175 (9th Cir.1989). This decision arrived between the Third Circuit’s Whelan decision and the Second Circuit’s Computer Associates decision. Johnson Controls is one of Oracle’s mainstays herein. In Johnson Controls, the claimant developed a system of computer programs to control wastewater treatment plants. The district court found that the structure, sequence and organization of the program was expression and granted a preliminary injunction even though the accused product did not have similar source or object code. Id. at 1174. Therefore, the standard of review on appeal was limited to abuse of discretion and clear error. Our court of appeals affirmed the preliminary injunction, stating that the claimant’s program was very sophisticated and each individual application was customized to the needs of the purchaser, indicating there may have been room for individualized expression in the accomplishment of common functions. Since there was some discretion and opportunity for creativity in the structure, the structure of the program was expression rather than an idea. Id. at 1175. Johnson Controls, however, did not elaborate on which particular structures deserved copyright