Archive for November, 2007

IFPI wants content-filtering

Posted in audiblemagic, censorship, copyright, eff, freedom, ifpi, mpaa, riaa with tags , , , , , , , on November 28, 2007 by qplqyer

Today I read that the IFPI, once again, is pressing the EU to make internet service providers (ISPs) block websites, protocols and content that infringes their copyrights. They propose three possible solutions, described in detail at http://www.eff.org/files/filenode/effeurope/ifpi_filtering_memo.pdf

I present to you the following breakdown which shows how each and every one of the so called “solutions that are not overly burdensome, nor expensive and that do not cause problems for regular service of ISPs” in fact only do those three things and do not solve the copyright-infringement problem in any way. It goes even further than that, in that these measures would actually block users from exercising the fair-use rights they were given in copyright law. Here are the details:

    Content filtering:

Content filtering, such as proposed by the IFPI, would be implemented by placing a device in the network of the ISP that checks all traffic for infringing content.

First, I would like to make some technical objections. The backbone of an ISP obviously has a bandwidth that has to be much bigger than those provided to their customers. They can generate these speeds through the way the internet (and networks in general) work: when deciding where some piece of information on the network (a packet) has to go, only a small part of this packet (the header) has to be looked at. What lies deeper in this packet is not necessary for sending the packet to the right computer, but is of use to the receiving computer (this contains the real data and is much bigger than the header). What the IFPI proposes is to insert a device in this network that also looks at the data contained in the packet, instead of only looking at the header. It is important to note that looking at the data in the packet takes up a lot more time than only looking at the header of a packet (which is what a switch and a router does). This does not impact network speed, since these devices will not be responsible for sending the packets to their destination. These devices do however insert data into the network, which might have an impact on the speed of the network.

Furthermore, data is split up in different packets, which means that only a small part of a song will be contained in one packet. If any device would reliably want to conclude that a certain song is being downloaded, it would have to keep track of a number of packets that all come from a certain server and go to a certain computer (this is called a stream). This would mean keeping track of every connection of every client of the ISP, something which is impossible to store in memory due to the amount of connections at any given time, and thus storing parts of these connections on a disk to compare with the previously captured traffic. Due to disks being very slow, and especially much slower than network traffic, it does not seem possible that any device might reliably inspect the traffic, since the slow disk will always hinder the speed of the copyright check. This is not the only problem with speed: searching a database can take a lot of time, especially if it as big as a database with all copyrighted works that exist. After a full stream with enough data is captured, this device would still have to search this database (again this operation is also limited to the speed of a disk!) to see if it matches any fingerprint.

What is even more so surprising is that current “solutions” for this problem, such as “Audible Magic”, come in small boxes, the same size as a server, without a large disk full of fingerprints, or a large disk for saving streams. This comes as a surprise because one would expect that, if a device really only blocks copyrighted content, the database of copyrighted works would be saved in the device. If this was not the case, and the database server would be on the internet, this means that every check the device has to make would generate traffic, causing the ISPs network to slow down.

So is it not possible at all? It might be possible, but it would mean setting up one of these devices at a lot of switches to make sure that the device can check all traffic. With a wet finger approach I would say these devices are capable of managing 1000 users at most (especially if these users regularly watch streaming content such as youtube movies), so an ISP with 1.000.000 customers would need something in the size of 1000 of these devices. At a cost of probably something in the areas of 2500$ (which is probably even an approximation on the low side) this would mean 2.500.000$ in installation costs. Costs that the users of this ISP will pay in the end. So we get to pay for being able to do less on the internet and to protect the interests of big media companies with enough money to lobby, sounds great, doesn’t it?

So we can conclude that technologically, there are a lot of questions that can be raised as to whether content filtering is possible without introducing a lot of costs. It can also be easily defeated by using encryption.

But there are even more objections, this time juridical in nature. We are given the rights in copyright law to do certain things with copyrighted works without needing the permission of the author, these are “fair-use” rights and include, among other things, the right to make parodies of protected works. Now, content-filtering devices do not see a difference between lawful use of a certain work and unlawful use. Therefore, by blocking everything that just checks out as sounding the same (it is acoustic fingerprinting after all), it does not seem impossible that these fair uses of works are blocked too. Especially if we do not know how the fingerprinting works (and if it does not flag too many false positives).

Furthermore, there is the issue of web-commerce and the downloading of legal songs, such as from itunes. If a webstore would want to distribute the songs one can legally purchase through torrent-files to reduce bandwidth-costs (and by this, reducing the price of songs), these solutions would block these legal downloads.


For the record: I know that solutions such as Audible Magic are mainly targeted against P2P networks, but this does not mean that in the future the IFPI will not start proposing to block these downloads from websites as well. Of course, by limiting the protocols that must be checked for downloaded content, it becomes more technically feasible, but it would still introduce quite soms costs for any big internet service provider.

Protocol blocking:

Every protocol can be used legally. BitTorrent, one of the most contested protocols by the movie (and probably recording) industry, for example, is being used to save bandwidth costs for projects that do not have much money. The most cited example of this are linux distributions, which use the BitTorrent protocol to distribute 4.7GiB files, which would cost them lots if every person that wanted the distribution would download these files from their servers.

The question is one about freedom though, should we allow companies, with commercial motives, to determine what protocals are deemed appropriate and which ones are not, just because they lose revenue because of them? If even one legal use of a protocol is hindered by a block, imposed by commercial companies, we have lost freedom. The freedom to choose which technology we would like to use for a certain (legal) task. I do not think this is a price we have to pay.

URL Blocking:

The same arguments as above apply to this proposition. But Url blocking goes even further than protocol blocking since if we allow this, it means commercial companies get control over what information we can read. Press freedom is one of the most important things we have and we should never consider to give this up for the commercial interests of the same companies. We win nothing out of it and they gain everything. The examples they give are a clear example of this: thepiratebay only hosts torrent files, which are files that do not infringe copyright themselves, they just give directions on where you can find files that are shared. The illegality of torrenthosting is still being disputed, but these companies want to block these sites already because they think it should be illegal (even if no legislation exists in a country that they would want to impose this block upon). Furthermore, the site also contains texts with critical comments against the music and movie industry. By blocking this website, we lose access to these texts as well.

Allofmp3 is a russian site, which operated under the russian law and did not breach russian law. Blocking a website when it does not breach the law of the country where it is hosted, because these companies consider it illegal, is once again an example of how these companies would restrict our perfectly legal use of the internet.

What is most important, however, is that, such as with all censorship, these companies would be playing the judge and noone would be able to verify if their decision was correct, because the website would be blocked.

Blocking URLs is always censorship and censorship should never be taken lightly. If a censorship is proposed to protect the commercial incentives of companies, we should strongly oppose this and suggest that these companies take the regular route to try the persons that they consider are doing illegal things, instead of trying to play judges themselves.

Don’t let this happen. Write to your EU-representative or whomever you know that might have an influence.