How We Lost The Internet

Overview

• This paper examines the assumption that the internet's stateless, point-to-point datagram delivery service is sufficient for modern, asynchronous, and point-to-multipoint applications.
• The paper explains how this model led to the development of private infrastructure like Content Delivery Networks and distributed Cloud data centers.
• It suggests these private overlays may have hindered the "Open Data Networking" goals of early internet advocates.
• The paper proposes reconsidering the adequacy of the internet's stateless datagram service model to address this contradiction.

Plain English Explanation

The early internet was designed around a simple model of delivering data in small "datagram" packets from one point to another. This worked well for many early applications, but as the internet grew, new types of applications emerged that didn't fit this model well.

For example, content delivery and online services that need to reach a large global audience are inherently more asynchronous and need to send data to many users at once, not just point-to-point. The internet's basic datagram model couldn't adequately support these new applications.

To fill this gap, companies built private "overlay" networks like Content Delivery Networks and cloud data centers. While these private networks solved the technical problems, they also introduced new issues. They went against the original "Open Data Networking" vision for the internet, where it would be a shared public resource.

The paper suggests these private, profit-driven networks may have contributed to some of the negative impacts of today's most profitable internet applications, like social media and targeted advertising. To address this, the paper proposes re-evaluating the fundamental datagram model at the core of the internet's architecture.

Technical Explanation

The paper argues that the internet's original assumption - that a stateless, point-to-point datagram delivery service would be sufficient for all network applications - has been challenged by the rise of large-scale, asynchronous, and point-to-multipoint content and service delivery applications.

The authors explain how this mismatch between the internet's architectural assumptions and the needs of modern applications led to the development of private overlay infrastructures, specifically Content Delivery Networks and distributed Cloud data centers. These private networks were built to provide the scale, performance, and reliability that the public internet's datagram model could not.

However, the paper suggests that the reliance on these profit-driven, private overlays may have been an obstacle to achieving the "Open Data Networking" goals of early internet advocates. It argues that the contradiction between these initial goals and the commercial imperatives of hypergiant overlay operators is a possible reason for the negative societal impacts of their most profitable applications and monetization strategies.

To address this, the paper proposes reconsidering the adequacy of the internet's stateless datagram service model as a potential step in resolving this contradiction.

Critical Analysis

The paper raises important points about the limitations of the internet's original architectural assumptions and the consequences of the private overlays that emerged to address those limitations. However, it does not provide a detailed analysis of the specific technical or economic factors that drove the development of these private networks.

While the paper suggests that the reliance on profit-driven overlays may have been an obstacle to the "Open Data Networking" goals, it does not fully explore the complex trade-offs and incentives that led to their widespread adoption. Further research would be needed to more rigorously evaluate the impacts, both positive and negative, of these private infrastructures.

Additionally, the paper does not address potential alternative approaches or architectural changes that could have better supported the original vision for the internet while still meeting the evolving needs of applications and users. A more comprehensive exploration of possible solutions would strengthen the paper's recommendations.

Overall, the paper raises important questions about the internet's evolution and the tension between its founding principles and the realities of modern, large-scale applications. However, a more in-depth analysis and a broader consideration of potential paths forward would enhance the paper's impact and usefulness for the research community.

Conclusion

This paper challenges a core assumption that underpinned the development of the internet's architecture - that a stateless, point-to-point datagram delivery service would be sufficient for all network applications. It explains how the inability of this model to support modern, asynchronous, and point-to-multipoint content and service delivery led to the rise of private overlay infrastructures like Content Delivery Networks and Cloud data centers.

The paper suggests that this reliance on profit-driven private overlays may have been an obstacle to achieving the "Open Data Networking" goals of early internet advocates, and potentially contributed to the negative societal impacts of some of the internet's most profitable applications and monetization strategies.

To address this contradiction, the paper proposes reconsidering the adequacy of the internet's stateless datagram service model as a potential first step. While the paper raises important questions, further research is needed to more fully explore the technical, economic, and societal implications of this issue and potential solutions.