View Christian Sandvig
Rapporteur: Paul Goodman
Dr. Christian Sandvig is an Associate Professor of Communication, Media Studies, as well as Research Associate Professor at the Coordinated Science Laboratory of the University of Illinois at Urbana-Champaign. He is a researcher specializing in communication technology and public policy. Sandvig is also a Faculty Fellow at the Berkman Center for Internet & Society at Harvard University.
With an emphasis on large-scale internet content distribution, Christian Sandvig described the infrastructure required for mass content distribution, laid bare the financial significance of this distribution, and explored a handful of case studies.
Sandvig launched his talk by exploring the opportunities for mass content distribution in 1995. Sandvig used as an example “Channemals”, a failed cable television concept. Channemals was pitched as a 24/7 channel dedicated to animals; it was soundly rejected by the likes of Barry Diller, Rupert Murdoch, and Ted Turner. The concept was deemed “too weird” to be supported by a major network, but the pitchman was offered airtime in exchange for $17 million.
Fast forward to today: Urine-off sponsors the Puppy Channel, and there are plenty of other strange animal-related websites and content: I can has cheezburgers, YouTube videos, and so on. The lack of channels has historically put control of content distribution in the hands of a small elite; the internet is changing this, but truly large-scale content distribution remains expensive and undemocratic.
Sandvig introduced the term “infrastructure splintering”, coined by Graham and Marvin to describe the transition from infrastructure as a modernist ideal (roads, power, rail, water, which are all easy to quantify) to a much less well defined area of having and not having. Sandvig also introduced Davis’ criticism of the “fantasy of bootstrap capitalism”, suggesting that independent infrastructure is not as promising as some have indicated.
Sharing the throughput required to support a variety of content types, Sandvig demonstrated that video content in particular places very high pressure on internet infrastructure. Sandvig argued that there are two internets: – “the internet that serves the things that people want to see, and the internet that serves content that is very popular.” If you have an unpopular piece of content, you can use your internet service provider to distribute that content. If you have a very popular clip (a clip of Michael Phelps, for instance), you will need to use content distribution networks (CDNs) to distribute your content to the masses.
In the 1930s, when broadcasting was in its infancy, scholars saw the machinery of broadcasting as a “machine of culture”. Now, people are posting photos and videos of data centers around the world. There is a clear parallel between the industrialization of culture in 1930s and today, where there are giant facilities. However, there isn’t much scholarship around this topic.
Data storage and distribution is a major enterprise. For example, Google today has 36 data centers and each may cost $300 million. A variety of CDNs populate this marketplace, including Akamai, Amazon Web Services, and so on. Sandvig argued that these organizations are trying to create a parallel internet for paid content. Akamai serves content for Microsoft, Apple, Yahoo, Amazon, MySpace, Facebook, CNN, BBC, Google, and so on. They serve 10-20% of web traffic and 60% of all CDN traffic.
Most distribution agreements are private, but Amazon Web Services’ version (originally built exclusively for Amazon and now available commercially, with a few locations) posts prices ranging from $0.170 to 0.221 per 10TB outbound. This is big business: YouTube spent approximately $1 million a month on bandwidth in 2007, was accused of losing $1.4-1.6 million in 2009, and reportedly spends more than $1 per visitor in 2009.
Sandvig examined the common fascination with the “long tail” concept. Commentary around the long tail – which is used to describe the distribution of content, concepts, and so on – tends to focus on the shape of the curve. Some suggest that systems that are self-organizing will demonstrate many sources of information with some channels having major audiences and most having little. Clay Shirky and others suggest that the internet offers those sources with few viewers greater exposure. Benkler argues that a minimum number of sources on left is troubling.
Sandvig argued that there is no theoretical justification for the long tail. He asks, “how would you draw an ideal curve for TV? One to one? Very few sources? Straight line?” Instead, Sandvig believes we should be asking “how do you move up the long tail?” He suggests that there are things that should be viewed by a long audience, and things that don’t need to be. Thus, we should be paying attention not to the shape, but where the content is on the curve.
Sandvig introduced the BBC’s iPlayer as an example of the complexity of mass content distribution. As opposed to a private concern intent on maximizing shareholder value, the BBC is in an interesting position in that they have massive video archives which people are interested in seeing. Furthermore, those people believe they’ve already paid for that content through their tax dollars. Thus, the iPlayer. The BBC initially thought that a good way to distribute their video would be via online downloads. However, that content needs to run through an ISP’s network and the BBC began to worry about their bandwidth costs given they don’t have additional funding for this activity. They later settled on using peer-to-peer file sharing, but people reacted negatively because they didn’t want their iPlayer client sending video to other peoples’ computers via their ISP. Finally, the BBC decided to sign up for a CDN (Level3).
When we talk about things like the long tail, we emphasize the fact that you’re providing content that no one is looking at. Instead, we should be thinking about the content that everyone is looking at. Examples of massive distribution: content around Michael Jackson’s death, information around the inauguration of Barack Obama, etc.
Who pays for video that should be massively distributed, and cannot be “monetized”? Civil society, public broadcasters, and so on. Terms originally used to describe broadcasting and communications infrastructure (mass communication, gatekeeping, common carriage, interconnection) are now back, as if the internet were retro-fitted for broadcast model.
Discussion period:
Q: a common alternative to CDNs is P2P, but doesn’t that hurt the long end of the tail?
NBC’s new venture uses P2P, transferring more of the bandwidth costs to other networks. In real-time events, multi-casting is a good tool. At the end of the day, P2P doesn’t get you out of the problem of a lack of diversity.
Q: Is there data on when the content’s position on the long tail becomes profitable for companies?
The answer is never, since online video is largely unsupported by advertising. Some advertisers do use CPM, however, but it really depends on how much money it takes to run your business. Also, there will be situations where content is valuable to have up, but may not be supported by advertising at all.
Q: How do you approach and teach terms (common carriage, etc.) when there isn’t transparency in this sector or vast public technical knowledge?
You can find your way in by reverse-engineering things – hang around on message boards where people in this space hang around. Use LinkedIn message boards, for instance.
Q: Where is this industry going?
There is a lack of normative discourse around the future of content distribution; we seem to be drifting. The Internet of ten years from now looks like TV of 40 years ago. If you want to publish tings only to a few people, the tools will be there.
Q: Gendering of P2P or the gendering of YouTube? Who is uploading and downloading – which sex?
See Canadian scholar Jonathan Stern (MP3: the meaning of a format).