Measuring the Utility of the Internet

The paper Modeling the Adoption of Network Architectures attempts to define a simple user utility-based model for measuring the migration from an existing network architecture to a new one. The model is based primary around standalone utility, the benefits a single user or organization receives from using the Internet, and network effects, or the benefits a user or organization receives because other users and organizations use the same network architecture. However, the authors lament that they have no way of determining real world values for any of their model parameters. This prompted to think about the value that the current Internet provides to different groups of people and ask if it is even possible to provide an accurate numeric value for something as complex as user utility.

At its core, the Internet provides a protocol, IP, for hosts to send messages to each other, and two protocols, TCP and UDP, for sending arbitrary-sized chunks of data between hosts. To hardware, manufacturers, this provides value in that if their equipment can communicate with a IP-aware router, then it can communicate to the Internet. To software developers, on the other hand, this means that they can rely on underlying TCP/UDP operating system calls to write network-aware applications. For specialized applications like web servers, even higher-level abstractions exist for HTTP-based communication, and yet higher abstractions that attempt to hide the HTTP request/response cycle. At their core, they are all built on TCP/IP.

Most organizations, however, will use the Internet through a variety of applications running on a variety of hardware devices. Their internal networks will also have to respond to a multitude of disparate needs. Our university campus, for instance, most likely has many kinds of "middlebox" network technologies like web filters and caches, and network activity monitors to detect malware. And, of course, the users in an organization can differ vastly in their needs from one another. Some users may user Internet primarily for basic communication: email, chat, and web-conferencing. Home users, on the other hand, may use any of a number of applications ranging from streaming video to casual Flash games. As a final interesting note, the standalone utility of the Internet for any user is constantly changing as new applications are developed and the infrastructure evolves to provide more bandwidth.

While the Internet itself provides only one network effect, the ability for any connected, addressable host to communicate with any other connected, addressable host, there are a host of applications that also take advantage of network effects. Social media, for instance, notoriously benefit from increased numbers of users, as well as any other communication tool like email. E-commerce businesses like Amazon.com also see increased revenue with more Internet adoption. As a last example, there has recently been increased effort poured into collaboration tools like Google Office, Google Wave, and Etherpad, all of which provide a limited amount of network effect-based benefits to some users in that they really only benefit from networking with a few others. at a time.

After considering not only the large number of different ways different users can benefit from a network architecture but also the different ways in which individual users may use such a broad and general tool, it begins to seem very difficult to deduce a general numeric value to assign to "user utility" for all network architectures. Perhaps the biggest value from this research, though, simply comes from the fact that it may cause others to consider what will be required to making breaking changes to the Internet.