Friday, February 03, 2012
The Used CD Store Goes Online
On Monday, Judge Sullivan of the Southern District of New York will hear argument on a preliminary injunction motion in Capitol Records v. ReDigi, a copyright case that could be one of the sleeper hits of the season. ReDigi is engaged in the seemingly oxymoronic business of "pre-owned digital music" sales: it lets its customers sell their music files to each other. Capitol Records, unamused, thinks the whole thing is blatantly infringing and wants it shut down, NOW.
There are oodles of meaty copyright issues in the case -- including many that one would not think would still be unresolved at this late date. ReDigi is arguing that what it's doing is protected by first sale: just as with physical CDs, resale of legally purchased copies is legal. Capitol's counter is that no physical "copy" changes hands when a ReDigi user uploads a file and another user downloads it. This disagreement cuts to the heart of what first sale means and is for in this digital age. ReDigi is also making a quiver's worth of arguments about fair use (when users upload files that they then stream back to themselves), public performance (too painfuly technical to get into on a general-interest blog), and the responsibility of intermediaries for infringements initiated by users.
I'd like to dwell briefly on one particular argument that ReDigi is making: that what it is doing is fully protected under section 117 of the Copyright Act. That rarely-used section says it's not an infringement to make a copy of a "computer program" as "an essential step in the utilization of the computer program." In ReDigi's view, the "mp3" files that its users download from iTunes and then sell through ReDigi are "computer programs" that qualify for this defense. Capitol responds that in the ontology of the Copyright Act, MP3s are data ("sound recordings," to be precise), not programs.
I winced when I read these portions of the briefs.
In the first place, none of the files being transferred through ReDigi are MP3s. ReDigi only works with files downloaded from the iTunes Store, and the only format that iTunes sells in is AAC (Advanced Audio Coding), not MP3. It's a small detail, but the parties' agreement to a false "fact" virtually guarantees that their error will be enshrined in a judicial opinion, leading future lawyers and courts to think that any digital music file is an "MP3."
Worse still, the distinction that divides ReDigi and Capitol -- between programs and data -- is untenable. Even before there were actual computers, Alan Turing proved that there is no difference between program and data. In a brilliant 1936 paper, he showed that any computer program can be treated as the data input to another program. We could think of an MP3 as a bunch of "data" that is used as an input to a music player. Or we could think of the MP3 as a "program" that, when run correctly, produces sound as an output. Both views are correct -- which is to say, that to the extent that the Copyright Act distinguishes a "program" from any other information stored in a computer, it rests on a distinction that collapses if you push too hard on it. Whether ReDigi should be able to use this "essential step" defense, therefore, has to rest on a policy judgment that cannot be derived solely from the technical facts of what AAC files are and how they work. But again, since the parties agree that there is a technical distinction and that it matters, we can only hope that the court realizes they're both blowing smoke.
On the program/data distinction, couldn't you also say for similar reasons that there's no distinction between computer data and other kinds of data? You might think of the contents of a CD or even a cassette tape as a program that, when run correctly, produce sound too...
Posted by: Paul Gowder | Feb 4, 2012 1:32:05 AM
Thanks for this great preview of the case. But I don't quite get your arguments about the non-distinction between programs and data. Even if all programs are data, as you characterize the Turing paper as demonstrating, it doesn't obviously follow that all data are programs. Could you say more about why the "if X is data, then X is a program" conditional is true or about your particular suggestion that MP3s are programs?
Also, does ReDigi place a lot of weight on the statutory definition from § 101 ("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.")? That seems like a pretty inclusive definition to me.
Posted by: Michael Kenneally | Feb 4, 2012 6:57:02 AM
Paul: the argument can also be made for a CD: it contains a "program" that runs on on a specialized processor and produces music. It probably doesn't reach the tape, which is encoded as analog data, rather than digital.
Michael, here's another a way of thinking about the equivalence. We're accustomed to thinking that a program P runs on data D to produce some output x.
P(D) = x
Turing developed the idea of a "universal" program U, which takes programs as input. What it does is simulate P, step-by-step, producing as its output the same result that P would have. To do that, it also needs access to the data that P would have run on.
U(P,D) = x
So we've just treated P as a program that runs on D, or as data that's input to U. But now we can reverse the logic. Not every program needs data as an input. We can think of D as a no-input program that produces x as its output.
D() = x
Of course, D only "runs" in the environment supplied by P. But we could similarly say that P only "runs" in the environment supplied by U or by the physical hardware. What is "program" and what is "data" is contextual and conventional, not something intrinsic to the information itself.
Posted by: James Grimmelmann | Feb 4, 2012 7:41:55 AM
I think I'm still missing something about your explanation, but now I wonder whether it really matters to the parties or their arguments whether the program/data distinction is "intrinsic to the information itself." A distinction can be "contextual and conventional" and yet still function well enough for legal purposes, right? One could describe the distinction between February and March as conventional and also contextual (since it changes every fourth year). That wouldn't mean we can't figure out whether a day between January and April is in February or in March.
Posted by: Michael Kenneally | Feb 4, 2012 9:37:45 AM
"it rests on a distinction that collapses if you push too hard on it." Isn't this true of pretty much *all* distinctions that the law makes? At least any distinction that is not purely binary is going to be subject to some version of the Sorites paradox. The relevant question, I think, to determine whether the distinction here is workable is not whether you can imagine a universal program, but whether such things are prevalent in the marketplace. They're not, so it is.
By the way, I initially thought you were going in an entirely different direction here, which is that the definition of "computer programs" in the Copyright Act itself seems incoherent (or maybe it just reflects a deep understanding of the computer science literature -- I hadn't considered that): "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." The use of "statements" and "indirectly" makes it sound to me like the definition of computer program might be sweeping in data files, but perhaps "statement" is a term of art that would rule this reading out.
Posted by: Bruce Boyden | Feb 4, 2012 1:18:11 PM
James: why should the analog/digital distinction make any difference? There have been analog computers...
Posted by: Paul Gowder | Feb 4, 2012 2:15:49 PM
In terms of the First Sale issue, Aaron Perzanowski and I have written quite a bit about this in our forthcoming Minn. L. Rev. article on Copyright Exhaustion and the Personal Use Dilemma. (Many thanks to James for his comments on an early draft.)
While we're supportive of the fair use arguments, among others, exhaustion pretty much takes care of the whole case if applied correctly. Section 109 (and 117, to the extent you also read it as a form of statutory exhaustion) are very fact-specific circumstances in which exhaustion applies, but the broader common law doctrine of exhaustion, as defined by Bobbs-Merrill v. Straus and Quanta v. LG, provides a clean mechanism for courts to allow alienation of digital copies by copy owners as well as any incidental or necessary reproductions to enable alienation.
Once individual alienation is held lawful, the direct and secondary infringement arguments against ReDigi fall by the wayside under CoStar v. Loopnet, Cartoon Network v. CSC, and Grokster and Sony v. Universal.
Posted by: Jason Schultz | Feb 4, 2012 2:16:14 PM
In terms of the First Sale argument, Aaron Perzanowski and I have written about this in our forthcoming Minn. L. Rev. article, Copyright Exhaustion and the Personal Use Dilemma.
While we are supportive of the fair use and other arguments at issue, we think exhaustion provides the court with a clean and concise approach to the problem. As copy owners, users are entitled to alienate those copies under the common law exhaustion rules handed down in Bobbs-Merrill v. Straus and Quanta v. LG. While Section 109 and 117 (to the extent one reads it as another form of statutory exhaustion) cover fact-specific instances, the broader exhaustion principle allows for alienation of copies in almost any context as long as copy rivalry is maintained. There is nothing in the Copyright Act or case law that limits exhaustion to analog copies, and we point to cases where even reproduction and adaptation are allowed to fulfill exhaustion's purposes.
Once alienation of pre-owned copies is held lawful, the direct and secondary claims against ReDigi should also fall under cases like CoStar v. LoopNet and Grokster.
Posted by: Jason Schultz | Feb 4, 2012 3:30:19 PM
James: Nice reminder on the ReDigi case. Any idea about typical turnaround for preliminary injunction motions like this one?
Jason: I really enjoy the articles you and Aaron have written on copyright exhaustion. I agree with the point that, for first sale to apply, the transfer of a digital copy has work like the transfer of a physical copy, and the seller can't keep a copy, or copies, after the sale. Copyright Exhaustion at 45.
If that's right, an important question is where should the burden of proof lie, regarding retention of copies. Do we require ReDigi to use a technology that scrubs extra copies from the seller's computer? Is there any way for ReDigi to verify that the seller hasn't burned a CD, or kept copies on an iPod or MP3 player? I think the answer to the first question should be yes, but the answer to the second is likely no. This is what makes me nervous in general about the analogy between physical copies and digital copies, which I explore in my article, A Second Look at the Right of First Publication.
If you think back to the analog days, there were certain indications that copy rivalry had been violated: a hand-crafted cover for the bootleg CDs you were selling, which you sell out of the back of a truck, etc. For secondary liability, we could look to indications like the fact that you are measuring the length of commercial albums and selling blank tapes of exactly that length to your customers for resale. See, e.g., A&M Records v. Abdallah, 948 F. Supp. 1449 (C.D. Cal. 1996).
In the world of the analog resale, it was easy to assume that consumers were getting rid of their only copy. Even when they could make a copy to tape, that seems to fall somewhere between handwriting notes from your favorite book and making an exact reproduction, due to analog degradation. I don't have any empirical evidence, but most of my anecdotal interactions with the digital generation suggest that keeping an extra copy would be the default position. I would like to think I'm wrong, but I think this danger should be taken into account when we try to make the jump from digital to analog, especially where digital resale runs counter to the text of the first sale exception in the Copyright Act.
Posted by: Jake Linford | Feb 4, 2012 4:24:14 PM
Paul, the distinction is that analog computers aren't programmable in software. To set them up for a particular problem, you have to adjust the value of resistors, make electrical connections, etc. by hand, in hardware. It's the fact that digital computers are programmable in software that makes them so powerful (Turing again) and also means that any given piece of digital information's meaning in a computer depends on its interpretation as program or as data.
Bruce, actually, universal programs are prevalent in the market: every personal computer ships with many of them. The operating system, the Java interpreter, the web browser, Flash if you have it installed -- all of these are universal programs, which take other programs as input. (Their formatting specifications vary, of course, but that's the case for any program that expects input in a particular form.)
As for the broader point, yes, ultimately any interpretive fact collapses if you push on it too much. A better way of putting my argument might be that this one collapses sooner than most. For example, the distinction between a "digital computer" and anything that is not one is, I think, more robust than the program/data distinction. Unlike, say, the line between literary works and musical works, the problem here is not that there are difficult border cases but that the categories of "program" and "data" are actually coextensive.
Posted by: James Grimmelmann | Feb 5, 2012 12:31:25 PM
"the categories of 'program' and 'data' are actually coextensive." I don't see this problem manifesting itself in daily use however. If you ask, "what programs do you have installed on your computer?", no one starts rattling off all of their word processing files.
Also, I'm not sure operating systems, etc. are universal programs in the sense of taking entire programs as data. Operating systems accept inputs and return outputs as designated by the API; but the entire program would not itself be an input. And I think you would need that for there to be any serious confusion. I don't see how the fact that certain programs exchange messages means that the data/program distinction is untenable, any more than the person/document distinction is.
Posted by: Bruce Boyden | Feb 5, 2012 12:59:49 PM
exec takes a program as input and executes it. Every operating system depends on something similar: without it, programs couldn't run at all. That's the first thing an operating system needs to do: APIs only make sense once there are running programs to call them. How does the OS on a typical computer run a program? By treating the program as data and copying it into RAM, a piece at a time, frequently modifying it as it goes (so that instructions in the program that refer to other parts of the program point to the right parts of memory). Then, and only then, is it possible to treat the program as program and execute it.
And I agree that conventional usage distinguishes "programs" from "data" and can draw a rough-and-ready line. But this would hardly be the first area where the law uses embraces conventional beliefs with questionable foundations. We're not quite in pi = 3 territory, but I am reminded of the enormous to-do over whether Windows and IE were "integrated."
Posted by: James Grimmelmann | Feb 5, 2012 2:42:42 PM
James, interesting point about exec and loading into RAM. But I can't shake the feeling that this is an arcane point the law does not need to worry about. One person's "pi=3" is apparently another person's "Samuel Huntington was the first president of the United States."
Posted by: Bruce Boyden | Feb 6, 2012 11:15:38 AM
Bruce, the technical question about the line separating data and programs might be arcane, though I tend to agree with James here, but the implications of that distinction are still pretty relevant. Once we've set up this distinction, we get very different treatment for programs and data when it comes to things like backup copies, essential step copies, and modifications. Why should the statute recognize the need for backups of programs, but not backups of data? That's a sensible question to ask even if data and programs are fundamentally distinct, and it's an even better question to ask if you think the distinction collapses.
Posted by: Aaron Perzanowski | Feb 6, 2012 11:59:00 AM
Thanks Aaron. I believe everything you said is consistent with everything I said. The tricky issue I see about a statutory exemption for backing up "data" is that all copyrighted works are data, meaning that the archival exemption would have far broader application (and lead to a much larger number of potential scenarios) than an exemption that applies to a class of works that is typically sold in a form requiring installation on another device to start with. But it's obviously a worthwhile question.
Posted by: Bruce Boyden | Feb 6, 2012 4:18:50 PM
For the people claiming that data and programs are distinguishable: into what category would you place programs in interpreted languages? Open a Python application on one computer, and it can execute a complex program. Open it on another computer (that doesn't have an interpreter installed) and it looks like ordinary quasi-English plaintext.
A similar thing happens with binaries across operating systems. If I open Firefox.exe on my Mac, it'll read it as a binary file (and probably interpret it as text.) Does something change from a "program" to "data" and back just because it ends up on different machines?
For that matter, the .app's that run on your Mac are actually disguised folders. There's an executable in there, but there're also images, plists, and tons of other forms of data. Should the law treat those different parts of the bundle differently?
Posted by: Andrew MacKie-Mason | Feb 14, 2012 7:09:55 PM