UNIVERSAL CITY STUDIOS, INC., PARAMOUNT PICTURES CORPORATION,. METRO-GOLDWYN-MAYER STUDIOS INC., TRISTAR PICTURES, INC., COLUMBIA PICTURES INDUSTRIES, INC., TIME WARNER ENTERTAINMENT CO., L.P., DISNEY ENTERPRISES, INC., and TWENTIETH CENTURY FOX FILM CORPORATION,
- against -
00 Civ. 0277 (LAK)
I, DAVID S. TOURETZKY, declare under penalty of perjury that the foregoing is true and correct:
1. I am currently a Senior Research Scientist in the Computer Science Department and the Center for the Neural Basis of Cognition at Carnegie Mellon University, at which institution I earned M.S. and Ph.D. degrees in Computer Science. I speak regularly around the world on such topics as cognitive science, artificial intelligence, robotics, and neural networks. I have authored three books, edited or co-edited nine collections of scholarly works, and authored or co-authored dozens of articles for scholarly journals, conference presentations, and the like. Over the past 25 years I have taught computer science material in a variety of formats, including brief tutorials at national conferences, week-long seminars for industrial clients, and semester-length university courses. My full Curriculum Vitae is attached hereto as Exhbit A.
2. I presently host a web page called the "Gallery of CSS Descramblers" on the Carnegie Mellon University web server (at http://www.cs.cmu.edu/~dst/DeCSS/Gallery). A printout of the Gallery is annexed hereto as Exhibit B. I created the Gallery in response to the Preliminary Injunction issued by this Court. The Gallery consists of a set of files containing source code, textual descriptions of algorithms, and discussion of programs that can decrypt data that has been encrypted with CSS, or that can recover the keys necessary for such decryption. One version of the decryption code has been published on a t-shirt; the Gallery includes an image of this shirt together with a link to the web site from which the shirt can be ordered. I do not link to the binary executable file for the program known as DeCSS.
3. It is my belief that source code is expressive speech meriting the full protection of the First Amendment. This belief results in part from my experience as a computer science educator. For example, my first book, "Lisp: A Gentle Introduction to Symbolic Computation" (Harper & Row, 1984), and its successor, "Common Lisp: A Gentle Introduction to Symbolic Computation" (Benjamin Cummings/Addison-Wesley, 1990), both contain several full-length programs that are meant to be studied and understood by the reader, not merely executed by a computer. Most texts on computer programming contain examples of this sort. In addition, in the graduate course I teach on artificial neural networks, students are required to read the source code for various neural net simulations in order to understand how the algorithms are implemented. My own education as a computer scientist benefited considerably from reading source code written by other programmers.
4. I am concerned that this Court issued an order prohibiting the defendants from posting source code for CSS decryption algorithms on the Internet. As a scientist, I feel it is imperative that anyone, not just academics, be allowed to participate in the ongoing analysis and improvement of encryption technologies. This requires a dialog among all interested parties, who must be free to engage in their own testing and reverse engineering, and to openly discuss the algorithms and their implementation. A normal part of such discussions among computer scientists is the sharing of source code. It is clear to me that some of the web sites presently posting and/or linking to DeCSS or the css-auth decryption routines are primarily part of the Internet-based scientific/intellectual dialogue described above, especially those sites operated by developers of DVD applications, or by cryptographers. Other postings/linkings appear to be made in support of the right of the defendants in this case to post the source code.
5. My Gallery is a combination of scientific dialog and political statement. In comparing the "anonymous C source code" for CSS decryption with the code in css_descramble.c (part of the css-auth package), I discuss certain structural and strategic differences between the two implementations of the decryption routine. Without access to the source code, I would never have been able to discern these differences or to present them to the world for commentary online. As another example, my web site contains a copy of a textual description of the CSS decryption algorithm by the cryptographer Frank Stevenson. How is one to determine whether Stevenson's description is accurate? The only reliable way is to compare it with the source code for an implementation that is known to be correct because it has been compiled and run successfully. My own initial introduction to encryption algorithms based on linear feedback shift registers came from reading the css-auth source code in conjunction with Stevenson's writings.
6. Besides the technical discussion of CSS encryption, a second purpose of my Gallery is to illustrate the futility of trying to differentiate executable source code from other forms of written expression whose First Amendment protection is already well established. To do this I provide several descriptions of the CSS descrambling algorithm. One is the source code for css_descramble.c, a computer program written in the C programming language. My understanding is that the Court has enjoined the defendants from posting this source code on their web site on the grounds that it could potentially be compiled and executed. Another Gallery item is a version of the same program written in a different programming language for which there does not exist a compiler. This language, invented by me, is similar to C, but contains sufficient syntactic differences that programs written in it cannot be handled by a C compiler. If the Court's intent is to enjoin publication of source code that is executable, then this version is not subject to the injunction. But if someone were to write a compiler for this language, they could unilaterally change the legal status of my file. Another item in the Gallery is a line-by-line translation of the C source code into ordinary English. Since this Court has stated that it does not wish to enjoin public discussion of the decryption algorithm, only the distribution of executable code, I believe that the defendants in this case could post my English translation of the algorithm without violating the Court's injunction. Of course, anyone skilled in C programming could turn the English version back into C code with little effort. But for those who are uncertain of their translation skills, I also provide as a separate Gallery item a version of the English description of the algorithm where each sentence is annotated with the equivalent C statement. Although this annotated version does contain executable code, it cannot be processed by a C compiler because that code is intermixed with English text and HTML formatting commands. The conclusion I hope will be drawn by visitors to the Gallery is that even when code is primarily intended to be executed, it retains expressive content, and that content can be preserved in trivially-derived variant versions that are not executable. Therefore, the executability of code is not an essential characteristic differentiating it from other forms of speech.
7. I have examined several dozen mirror sites containing DeCSS and/or the css-auth source code. None of the sites I examined advocate the piracy of DVD movies. Rather, they appear to have been created as a reaction to the plaintiffs' attempt to suppress distribution of a computer program which the plaintiffs do not own (DeCSS), and to suppress the distribution of source code (css-auth) which most computer scientists view as speech. Thus I believe the mirror sites are making a political statement, even when the prohibited files are posted without additional commentary.
Dated: April 27, 2000
(signature: David S. Touretzky)