Images source: Wikimedia Commons, Matt Brim

The October issue of Discovery magazine has an article that piqued my interest, entitled, “This Man Wants to Control the Internet. And you should let him.” The man is Caltech professor, John Doyle, an expert in control theory. His field models dynamic physical systems, which includes things from a mechanical heart to space flight. The key idea is achieving a desired or steady state for one of these systems by taking current information about its state, and “feedback” that information to the system to make adjustments. These feedback system are mathematically modeled. When the system is non-linear and dynamic, for instance a airplane flying through wind currents, the mathematics required become quite sophisticated.

Doyle and his collaborator and fellow CalTech professor, Steven Low, have developed an improved protocol over TCP (or Transmission Control Protocol.) TCP describes how packets of data should be delivered and received over the Internet. FTP, email and WWW applications all rely on TCP. Using control theory, their protocol, FastTCP^TM, clocks the time a data packet takes to get to a final destination and make adjustments to optimize its stream of packets. Standard TCP does not take this extra information into account, and relies mostly on a strategy of monitoring lost packets. That is, packets that don’t make it to the finally destination. In the 2006 Supercomputing Network Bandwidth Challenge, they won it with a maximum throughput of 17 gigabits (a full-length movie) per second.

Improvements to the Standard TCP will be important in the coming years, as multimedia services (such as movies on demand) will increase the demand of the current network. Already, VOIP services do not use TCP, because packets sent using TCP cannot be received and sequenced fast enough for real time applications like phone calls.

Doyle and Low, along with Cheng Jin formed the startup, FastSoft, to sell products based on FastTCP^TM. However, they have trademarked their name and have submitted patents their technology. This is an important departure from the origins of the Internet, as no one owns that Standard TCP. Having to license or buy FastTCP^TM from FastSoft has implications to the future of the Internet, which could lead to its fragmentation.

Last month, at team from Indiana Univeristy, the Technische Universitaet Dresden, Rochester Institute of Technology, Oak Ridge National Laboratory and the Pittsburgh Supercomputing Center won the 2007 challenge. They achieved a peak transfer rate of 18.21 Gigabits/second and a sustained transfer rate of 16.2 Gigabits/second. It is not clear to me what kind of IP, the team from IU has on their technology. However, the received funding from the NSF, which may mean place of some or all of their research into the public domain.

Demands for bandwidth are only increasing. A complete overhaul of TCP is years ago, and involves incremental change, because the network at stake (that is, the Internet) is so important, which Doyle explain the Discover article. How we meet those demands is already controversial.

Susan Crawford notes that Comcast is already traffic shaping bits, by flagging packets by people using BitTorrent. (She also has a nice description of TCP in this post.) Meeting this growing need, the network can improve performance in various ways including: upgrading the infrastructure, such as laying fiber optic cable; improving data compression algorithms, and improving the protocols that control data traffic. In all these areas, the ownership and regulations of these technologies have huge implications on accessibility and adoption of the Internet. Although the Discover article’s title “this man wants to control the Internet” is a play on Doyle’s field of study, it raises an important point. Having public and private protocols may not only make parts of the inaccessible to each other, but further increase bandwidth as another form of economic inequality.

I’ve been slowing making my way through a very good book “Innovation and Incentives,” by Suzanne Scotchmer from UC Berkeley. I’ll close with a quote from her chapter on “Networks and Network Effects”:

“The protocols of the Internet and worldwide web were developed at public expense and put into the public domain. Given what turned out to be at stake, that is probably one of the most fortunate accidents in industrial history.”

Image source: IEEE

Noah Brier posted an interesting link recently to an article claiming that Metcalfe’s Law, which famously has been paraphrased to be the value of a network exponentially quadratically grows with each additional node, is wrong. This is a provocative thesis because the law is widely trumpeted in all that is good about the Internet.

In the theoretical long term view, networks with more nodes will encourage late adopters to join (and pay.) People often cite the instance of email, the rates of people setting up (and paying for) email accounts in the 1990s rose much faster after it was possible to communication with people outside a user’s ISP, being AOL, Compuserve, Prodigy, or an academic institution. However, network adoptions plays out differently in other cases.

The disconnect here may be that in the definition of value, more specifically value by whom. In the valuation of a network, the network of 100 is worth more than a network of 10. According to Metcalfe, it’s worth 100 times (10^2) the smaller network. For the consumer, it is clear that the value of the telecommunications network grows with each new person, especially if the costs for the user doesn’t increase. However, consider combining access of two competing networks from the perspective of the network operators. From this angle, the smaller network will clearly see more benefit than that larger network, as the article suggests. Further, both operators will see an increase in the costs associated with network traffic (from hardware to customer service,) with no new paying customers. If the operators are charging an unlimited usage prices, there is little upside to combining networks, which shows why network operators tend to resist interoperability. Why would a market leading operator take on additional cost with increased revenue and help a smaller competitor?

Coming back to the definition of value, the value of the network actually doesn’t change if the willing to pay (i.e. the ability to extract fees) from the customers of the network. I’m still trying to grapple with the difference between the value of the network from the customer versus the operator. That is, the problem is that as although the value of a larger network may grow, there willingness to pay by the customer does not. In fact, it may even shrink, which was noted in a MobileCampNYC talk by some smart folks at Air Arts last Saturday. Consumers have been conditioned to expect prices to fall, especially in the area of telecommunication services, even as nodes (and the value of the network) increases. Is the reason that customers don’t want to pay for that additional nodes because the new nodes are less valuable to the customer as the article suggests? Is it a matter of marketing conditioning from other services industries associated with the properties of the economics of scale that prices for services should go down over time?

Although the article was published last year, the valuations for networks will only receive more scrutiny after Facebook’s USD$15 billion valuation, which is about 25 times what News Corp paid for MySpace, which has more than double the users than Facebook. Expect more discussion to follow.

Here is a photo I took earlier this year at the then brand new Bangkok International Suvarnabhumi Airport (BKK). I was thinking about the image today, while doing some reading on social network theory. Airports are transportation hubs and feel weird when they are empty, because they are not acting out their function as place of transit. Shininess didn’t help matters at all either.

Some new points have instant connections while others don’t. I’m interested in how new points form and get connected into a network. Some points appear with a social network in place, say a baby born into a large family. Other points have connections which exist in other networks and their entry into a new network gives them instant connections as well. Steven Colbert made 1 million facebook friends in a week, which shows spill over from other networks. This is all obvious, but I’m curious if there are ways to show how different networks interrelated.

I will add more here, but I’m late for dinner.

Image source: usgs.gov

My last post to flowtv.org described the work by Kevin Werbach, a legal professor at the UPenn’s Wharton School of Business. I first heard about him at this year’s Telecommunications Policy Research Conference. He is looking at how the different forces pull the Internet together as well as pushes them apart. I wrote about how it got me thinking about how the Internet is fractal, and how important is it to have models like Werbach’s to help explain it.

At first, because the Internet works so well as a decentralized network, Werbach’s suggestion of the idea of a fragmented network comprised of archipelagos and walled gardens seems unlikely and unwanted. However, Techcrunch is reporting that in China, attempt to access Google and Yahoo are getting redirected to the homegrown (and approved) Baidu. A chance of this kind of fragmentation is quite real, which could also mean that the ICANN testing of non-Roman language domain names might be too little, too late.

So, I downloaded the album, which I paid $USD6. Frankly, for what is being reported as the death of the record label, it was a little anti-climatic. That is, I paid for a album over the internet that I could have gotten free. The album is good, sort of what we’ve come to expect from RH. Although, it usually takes me a few listening sessions for me to figure out what is going on in their albums. Now what?

For the second act, Trent Reznor announced that NIN has fulfilled their music label contractual obligations are now planning to go it alone, which isn’t all that surprising.

Starting your own record label or self-publishing is not a new thing, as seen with Ani DiFranco’s Righteous Babe started in the 1990, to cite one example. So, the real question still to be answered is, can this scale outside established musicians who admittedly benefited from aspects of the traditional music industry?

Will the home grown bands who grow a fan base through touring, myspace, pitchfork reivews and other grassroots efforts continue on a DIY career path or will they eventual jump to labels, as did Clap Your Hands Say Yeah?

Image source: In Rainbows

Way back in 1999, Public Enemy released “There’s a Poison Goin’ On” only on the internet, with indie label Atomic Pop. After sluggish sales, they eventually sold a CD version as well. I applauded PE for their efforts, although it wasn’t surprising with the outcome. Broadband penetration was much lower and PE has past it’s peak of popularity. Eight years later, the internet is buzzing with Radiohead’s announcement that they are releasing their latest record, In Rainbows, without a label. You can buy the disc set at a pricey £40.00 ($USD 80) or pay whatever you want for the download.

Of course, Radiohead is in a better position to do sometime like this kind of experiment, after having sold millions of records and toured extensively throughout the world. I’m sure the die- hard fans (of which there are many) will purchase the disc set. However, after Thom Yorke’s last album was circulating the internet months before its official release, Radiohead figures to try to recoup some of the lost revenue from p2p file sharing.

I paid £2.50. There is a £0.45 credit card process fee, which they only tell you about at the end of your purchase. I’ll give them a “free pass” on that one.

I got the download code, and will be able to get the album starting on October 10th. I’ll probably wait a few days, because I’m sure that their servers will be clogged at first. I’m not in a rush, as my Radiohead interest peaked a few years ago (I bought Amnesiac on the day it was released.)

I would love to see the numbers, and the distribution of what people’s willingness to pay for the album. It isn’t clear if this is sustainable for other bands, even famous ones. Radiohead is getting a lot of free press for being the first. The 10th band who tries this, won’t have the added benefit of extra publicity. I mean, I’m writing about buying an album on the internet.

I’m sure it will be available on the internet for free the day it is released. I could have gotten activation code for free as well, but I want to reward them for their efforts. I’m also buying the album for $USD5, which is actually how much it is actually worth to me. Although, it may be not be “rational” in the purely short term economic sense, I definitely wanted to reward Radiohead for working in this way.