During current debates over Internet governance, many commentators warn that the Internet could fragment in ways that would cause the benefits associated with the Internet to dissipate. Typically, these arguments reflect an absolutist character that inexorably leads to universal interconnection and interoperability. Although connectivity and standardization yield clear benefits, a review of the existing architecture reveals many circumstances in which the real-world networks have opted in favour of some degree of fragmentation.
This paper describes current examples of fragmentation in the Internet’s physical architecture, address space and protocols, and in the legal principles governing the Internet. It then advances analytical principles, such as diminishing marginal returns, heterogeneity in valuation, the lack of unique value in pairwise potential connections, and non-linear increases in cost, that can serve as heuristics for identifying the circumstances in which fragmentation is more likely to be either beneficial or detrimental. Finally, the paper identifies alternative institutional forms, such as gateways and arbitration, that can mitigate some of the problems associated with fragmentation.