Book notes: "Internet architecture and innovation" by Barbara Van Schewick

I took some notes while reading "internet Architecture and Innovation" of Barbara Van Schewick, director of the Stanford Center for Internet and Society. They are random but I thought some might find it useful...

 

This book is highly recommended for lawyers, economists and professionals in public policy that work in Internet issues. 

The aim of the book is to analyze how the architecture of the Internet affects the economic environment for innovation and the public policy implemented for its development. 

The book is divided in four parts. The first one, called Foundations, lays the framework of the discussion by explaining how the architecture of any system, and the Internet in particular, constrains the behaviour of the users of the system. The architecture not only imposes technical restrictions to certain actions, but also create important economic incentives to do or not take other actions. 

The second part of the book -"The End-to-End Arguments and the Original Architecture of the Internet"- describes the basic architecture of the Internet and compares its advantages and disadvantages to other possible designs. Van Schewick starts by comparing modular designs, in which components are loosely coupled, with integrated designs, in which such coupling is tight. Modular designs are less complex, allow for independent component design and higher modifiability, but imposes restrictions to the global performance of the system (as the overall structure of the system is not easily modifiable). 

Next the author explains shy layering is a special form of modularity. Layering constrains how modules interact. In a computer, for example, the hardware (hard disks, monitors, etc) are the lower layer, followed by the operating systems that provide application programming interfaces (APIs) to applications, which are in the highest layer. 

With this Schewick is able to explain one of the major insights of the book, which is the End-to-end arguments. The e2e arguments guide the placement of functionalities in a multi-layered system and provide the rationale for moving application specific functionalities upward. Here Van Schewick demonstrates how there are two versions of the e2e argument, a "broad" and the "narrow" version. The narrow version was presented in the first paper by Saltzer, Reed and Clark: "A function should only be implemented in a lower layer, if it can be completely and correctly implemented at that layer". The broad version was described in a paper years later: "A function or service should be carried out within a network layer only if is needed by all clients of that layer, and it can be completely implemented in that layer". 

There are important differences between these two versions. While the narrow one only applies to e2e functions, the broad version applies to the system's complete functionality. The narrow version distinguishes between partial and complete implementations of e2e functions. 

The consequences of the difference are also important. The Net Neutrality principle is derived from the broad version, but does not necessarily derives from the narrow version. 

The Internet is a modular relaxed layering with a portability layer. It consists of four layers: Application, transport, Internet and link. The Internet layers acts as the portability layer, since the higher layers can communicate with it but not with lower layers. 

The third part is dedicated to demonstrate how modular architectures impose less restrictions in innovation than integrated designs. Changes in integrated designs generate ripple effects, which means that a change in a particular aspect of the system usually creates the need to change other parts of the system. This does not happen in modular architectures, since the components of the system only have access to the visible information of a particular module and cannot interact with the hidden information. Under uncertainty, modular architectures provide more incentives for innovation than integrated designs because it allows more options of particular modules to be developed by independent developers. Modular architectures also reduce the costs of adaptation of the system to the changes in architecture. 

With respect to the firm's costs, e2e designs reduce transaction costs greatly and do not impose a heavy burden on coordination costs, which may be the particular advantages of a firm. Integrated systems create competition between proprietary systems while modular architectures create competition between components with the same architecture. 

The book continues analyzing the effect of architecture in competition. Van Schewick tries to prove how the owner of a totally controlled network, which is able to discriminate content, has such an incentive to discriminate, as opposed to a e2e architecture where no one is able to discriminate the different that is communicated through the Internet. Independent developers have real incentives to innovate in e2e architectures that they do not have in totally controlled architectures. 

This is true also from a public policy perspective. Descentralized environments allow for more easier entry of disruptive technologies and for a more active participation of venture capital, for example.