Peer-to-peer and Exponential Math
Fred Von Lohmann criticizes Solveig Singleton’s new essay for failing to discuss the darknet critique of DRM. Ray [Gifford, I assume] at PFF had this reaction:
The talismanic authority of the Darknet paper baffles me. It simply proves too much; namely that because some can circumvent DRM and, in a sense IPR, therefore there should be no DRM. It is not a completely bankrupt argument–you can certainly argue that the net effect of, say, antitrust laws or drug laws is negative. With Darknet, though, it is treated at something of an instant QED. To the contrary, it simply makes the point that a black market will arise on the internet for illicitly copied content. The challenge of law (and the usefulness of property rights) is, through social norms and legal sanctions, to shrink the size of that black market so that productive activity continues and a market thrives. Seems to me, on balance, that’s what the DMCA is doing in its best instantiations.
I think this response misunderstands the darknet critique, and the nature of peer-to-peer file distribution more generally.
Consider the analogy to the drug war. If the United States government stops 90 percent of the cocaine coming across the US border (and if we assume that not much cocaine is produced domestically), that will–all else being equal–reduce cocaine consumption by about 90 percent. We can argue about whether that’s good policy (I think it’s not), but the basic mechanics of the situation are straightforward: reduce the amount of drugs coming across the border and you reduce domestic consumption.
The “black market” in peer-to-peer song trading is very different. Let’s suppose we lived in a world without DRM. A new album comes out, and 1000 people purchase it and upload their copies to a peer-to-peer network. Within a few days, a million people download the album.
How long will it take those million people to get the file? An important point here is that the distribution of files will follow an exponential curve, at least until you get close to the saturation point (at which point it’ll level off into an S-shaped curve). The first x people will each share their copies with one other person, causing 2x people to have it. Then each of them share a copy with another person, causing 4x people to have it. The time it takes to distribute the files to all one million people depends on the doubling time. If we assume that the doubling time is 6 hours, then it will take about 3 days for the file to reach a million people.
Now suppose we introduce DRM into the equation. Let’s imagine that DRM is extremely effective: say, 95 percent. So now of the 1000 people who would like to share their copies, only 50 have the technical savvy necessary to download a cracking tool. How does that effect the distribution of the files? Now we have to grow from 50 copies to a million. If we again assume a doubling time of 6 hours, that will now take a little over 4 days.
So let’s ratchet up our war on file sharing by introducing the DMCA. Now, not only are songs DRMed, but we’ve banned circumvention tools. Let’s say that this reduces the use of circumvention tools by 90 percent. So now there are only 5 people in the world who have a physical copy of the song and are willing and able to find a (now illegal) circumvention tool and upload copies to the Internet. So now we have to grow from 5 copies to a million. How long will that take? If the doubling time is still 6 hours, it’ll take about 5 days.
Now, I’m deliberately choosing very conservative numbers. A single peer-to-peer client can distribute a lot more than one copy of a song in 6 hours, so a more realistic time for music is probably measured in minutes, not hours. (It might be about right for movies, although the doubling time will surely drop as bandwidth increases) It’s likely that more than 5 percent of uploaders would be able to find and download a circumvention program if such programs were legal. And it’s certainly the case that the DMCA is less than 90 percent effective at stopping the spread of circumvention tools to those who seek them out.
So even if we stop 99.5 percent of illicit files “at the border” (to return to our drug war analogy), that won’t have any perceptible effect on domestic consumption of those files. At most, it will delay the availability of those files for a day or two. I suppose it’s possible that a tiny number of hard-core fans will be deterred by their inability to get their music on the day of release, but they’re probably people who would buy the album for the cover art and liner notes anyway. The vast majority of illicit file-traders will simply wait the extra day.
In short, DRM simply doesn’t “shrink the size of that black market.” This isn’t a matter of DMCA oppoents unreasonably demanding perfection. Our point is that DRM doesn’t have any noticable effects on peer-to-peer file sharing at all. This particular emperor doesn’t have any clothes.
Moreover, it’s worth noting that the first leak of a song or movie often happens even before the release of the commercial version. In that case, DRM might not even delay the spread of the file, since the it would be widely available already the day the first DRMed copy hits the streets. Here is an article from 2002, showing that bootleg copies of movies were showing up on peer-to-peer networks before they’d even hit theaters, and months before the release of the DVD.
Patrick Ross adds:
Solveig, I believe, makes clear in her paper the goal of reduction, not elimination, of black markets. If critics continue to ignore that point there’s not much hope of constructive dialogue.
I think a good starting point for a constructive dialog would be for Ross or one of his colleagues to explain how DRM accomplishes a reduction of black markets in infringing copies (i.e. “cyberspace infringement,” as I described it in yesterday’s post). I’ve yet to see any explanation of how that’s supposed to work. Simply asserting over and over again that we need DRM to reduce peer-to-peer file sharing doesn’t demonstrate that DRM actually can reduce it.
-
Commons Music
-
Steve R.
-
Tim
-
amonynous
-
Tim
-
Steve R.
-
Tim
-
amonynous
