The Policy Implications of End to End: A Workshop at Stanford Law School

e2e Map, by Yochai Benkler | back

The folloing maps some of the policy tradeoffs suggested by the end-to-end conference

At the highest level, there is a tradeoff between

          FREEDOM V.           CONTROL

Which is played out in both economic and political domains

          POLITICAL
First party v. third party [1]
Voluntary v. involuntary [3]
Universality v. balkanization
Noncommercial v. commercial [6]
Autonomy v. order [7]
Popular democracy v. elitist democracy [8]
          ECONOMIC
Innovation v. manageability [2]
Innovation/growth v. allocation [4]
Network externalities/social value v. private returns to investment [5]

        .............................................

Other notable points from the discussion:

The interrelated effects of network-based solutions to local problems that might be thought of separately as independent matters of social, economic, or engineering concern, but in fact affect a much broader set of concerns because of their implementation in network design.

The global effects of network design and decisions to shift from end-to-end suggests that the cost/benefit analysis of specific solutions to specific problems must take into consideration the side effects of changes in network design across all the domains it affects.

Policy preferences for competition or freedom may require iterative application of the end-to-end argument at each layer of the communications infrastructure vis-à-vis higher layers. The simplest instance is the capacity of control over a lower layer to stymie competition at higher layers. OS implementing application specific functions to prevent competition in applications is an intuitively simple example. What the conference underscored was the possibility in general of departures from end-to-end arguments to permit those who depart from end-to-end to interject measures of control that can have either economic or political implications, and that can affect any layers above the departing layer. The general answer to Jerome Slatzer's question regarding what is the appropriate layer at which to implement end-to-end, then, is a question that depends on the values that lead one to adopt end-to-end arguments. To the extent one supports end-to-end design for reasons of opening economic competition or decentralizing political control over information flows, end-to-end arguments must be deployed at each layer vis-à-vis all higher layers in order to eliminate points of control that could undermine competition or the free flow of information at the highest layers.

back to "about" page

¹Question is whether communications are controlled solely by first parties, or whether third parties can intervene to prevent or otherwise shape the content or structure of the conversation.

²Here we have issues of trust and firewalls, for example, versus ease and speed of deployment of new technologies.

³Whether what shapes a communication is voluntarily used by the participants of the communications vs. whether the control mechanism is involuntary (e.g. filters, firewalls, caching, mirroring).

⁴Innovation v. manageability reenacts an old dichotomy in the arguments about the sources of wealth, in particular between growth through innovation and efficiency through allocation of existing resources. QoS, for example, suggests such a tradeoff, as does caching, where investment in efficient allocation attenuates the need for investment in provisioning, where investment in provisioning could remain neutral as among types of end uses but price-based allocation is not.

⁵This raises the question of possible divergence of private returns from social returns to various patterns of behavior. E.g., an ISP that speeds up delivery of its proprietary information relative to its competitors' may see a private gains in getting more of the traffic of a smaller universe of users, even though tilting its network so will degrade the overall value of the network to its users and make the communications network lumpy, losing some positive network externalities.

⁶In numerous instances, caching. mirroring, QoS, one sees a situation where the focus on security, manageability, and allocation take the form of solutions that make the network more amenable to commercial communications that can directly pay their freight, as opposed to communications that are done on a nonprofit or amateur basis. This, in turn, evokes questions about whether the mix of information and cultural production in society is better when different types of producers produce information.

⁷This is seen very effectively in the security-concerns based segmentation of the Net, as with firewalls.

⁸This juxtaposes a notion of democracy based on popular, end user to end user interaction, as opposed to democracy mediated by information production elites. Akamai mirroring CNN is an example of network design that advantages high-production value information communicated by an element of the "elite" over lower production value discussion forums that are not mirrored.

Stanford Law School home