Skip to main content

Science: Fixing the internet


This article, published in Science by Jonathan Zittrain, outlines how the web has changed from its initial concept to its current structure. He first outlines the overarching initial design. First, the “procrastination principle”: essentially “set it and forget it”, meaning components of the web were made freely available, and the designers stepped back and let the world figure out what to do with it. Secondly, the web’s layered architecture: its protocols were designed to ignore what they depended on and what depended on them. This means, for example, the packet switching protocol (a very low level component to the web) does not care what it is being used for: you can send photos, videos, text, et cetera, and whatever you use it for is irrelevant to the packet switching protocol. Conversely, a user of the web can send an email and not have to concern themselves with what happens behind the curtain. Finally, the core concept of decentralization specifies the web’s structure will not depend on any central authority, and instead be designed with equity in mind: a webpage made in Romania will be accessible by a user in the United Kingdom simply by design, and no user will need to ask a gatekeeper of the web for permission to view the page.

Zittrain then moves into the modern structure of the web, which in many ways negates the core concepts the web was originally built upon. He is primarily concerned with centralization and the destruction of the web’s layered design. Massive web hosting services, namely Amazon Web Services (AWS), control a massive amount of the web’s traffic nowadays. As of 2017, Amazon accounted for around 40% of the cloud hosting business. As a result, surfing the web “can now mean simply jumping among Amazon Web Services’ hosting servers”. There is now a central authority overseeing a vast component of the web. The layered design of the web is crumbling as well. Rather than building a website using the protocols themselves, many utilize services and platforms like Facebook or Google that contain the entire project under one umbrella.

In lecture, we discussed the web’s structure as a giant strongly connected component with millions of nodes pointing in and nodes pointing out (the bowtie structure). This graph is generated by following links and seeing where they point. However, if one were to draw such a network and separate the strongly connected component by host or provider, we would quickly see large clumps from AWS, Facebook, Google, etc. form, going against the framer’s original vision of the web: a strongly connected yet decentralized network of webpages. Many of the links within the component stay within a hosting provider and don’t actually hop to a different part of the web, just a different part of the host.


Leave a Reply

Blogging Calendar

October 2019
« Sep   Nov »