Arguing with Harold is a great fun because he is so terrifically passionate and articulate as an advocate. For those who don't know Harold, imagine a one man army–think Rambo–being ready to take on the world. So call me crazy for joining his turf to tussle, but here goes.
Harold's perspective on network neutrality is clear and relies on a powerful premise–a vertically integrated firm cannot be trusted to allow competition at the applications layer. For Harold, then, the solution is either enforced modularity–confining the transport provider to pure connectivity–or a virtual modularity where the interfaces are standardized and open.
The problem with Harold's vision is two-fold–(1) it underestimates the considerable logic in favor of openness that will lead most companies to allow rival applications to ride on their platform (yes, there are exceptions to this principle, but I will get to that); and (2) it underestimates the possibility that some applications if not delivered on a vertically integrated basis will not be delivered at all. Under the Computer Inquiry regime Harold praises, for example, consumers did not get the benefit of voicemail services because it was not economic for independent providers to offer them and, for many years, the incumbent phone companies were prevented from offering it.
If one agrees with my critique of Harold's vision (the details are here), the next question is what type of regulatory regime should one adopt. My vision, as Harold notes, does not dismiss that broadband providers with market power may well have incentives to discriminate and undermine applications providers (particularly those that threaten their core revenues). It does require, however, that regulators find ways to focus on the competitive harms without adopting rules that will give rise to unintended consequences, foreseeable inefficiencies, and potentially pick winners and losers. As my most recent post notes, however, there are real questions as to whether the FCC is institutionally prepared to oversee an antitrust-like model of regulation.