This article by Rep. Rick Boucher (D, VA) on network neutrality, is really worth a read –
Recently, executives at some telephone companies have indicated that their business models for providing broadband service include not only charging their end-user customers for an Internet connection but also assessing a fee on websites for users to reach them more quickly. They claim that to offer advanced content such as multiple video-programming channels in competition with cable they need to prioritize their bits to deliver quality programs. They then propose that they will give the same priority access to other companies that pay them for it.
Essentially, what these executives are proposing is the creation of a two-lane Internet where larger, more established websites with financial resources could squeeze out smaller, emerging websites. One clear victim will be the innovation that has thrived on the open Internet. Startups simply could not afford to pay for fast-lane treatment nationwide. One must ask where the next Google or Yahoo will come from if new innovative companies can receive only inferior, slow-lane Internet access.
Internet2, a nonprofit partnership of universities, companies and affiliate organizations, including federal agencies and laboratories, has been studying this matter and has demonstrated that a multitrack Internet model is unnecessary to assure quality of service. Internet2 has for the past seven years deployed an advanced broadband network to more than 5 million users and has learned that in a network with enough bandwidth there is no congestion and no bits need preferential treatment because all of them arrive quickly enough to assure excellent quality, even if intermingled.
In countries such as Japan and Korea, network speeds over the last mile of 100 megabits per second (mbps) are common. In the United States, our typical speed is less than 1 mbps. If broadband providers would increase their network speeds to approximate those in other countries, all content would reach consumers with assured quality. No prioritization of bits would be needed.