A fundamental issue in the network neutrality debate is the extent to which network operators should be allowed to discriminate among Internet packet streams to selectively block, adjust quality of service, or adjust prices. This paper first reviews technology now available for traffic discrimination. It then shows how network operators can use this technology in ways that would make the Internet less valuable to Internet users, and why a network operator would have financial incentive to do this if, and only if, it has sufficient market power. A particular concern is that network operators could use discrimination to extract oligopoly rents from upstream markets that are highly competitive. This paper also shows how network operators can use the very same technology to discriminate in ways that benefit Internet users as well as the network operator. Thus, network neutrality supporters are right to fear unlimited discrimination in some cases, while network neutrality opponents are right to fear a policy that imposes strict limits on discrimination. From this, we argue that the network neutrality debate should be refocused on the search for a balanced policy, which is a policy that limits the more harmful discriminatory practices in markets where there is insufficient competition, with little interference to beneficial discrimination or innovation. We apply this balanced policy in a few controversial scenarios as examples. There has been too little attention on the possibility of a nuanced balanced policy, in part because the network neutrality debate is focusing on the wrong issues. This paper argues that the debate should shift toward the complex details of differentiating harmful discrimination from beneficial discrimination, and away from high-level secondary questions like whether discrimination is inherently just, who ought to pay for certain Internet services, how important general design principles are, what abstract rights and freedoms consumers and carriers deserve, or whether network operators can give their affiliates special treatment. Reality is more complex than these questions would imply, and none of them will serve as a basis for a sufficiently specific and effective policy.