It’s Time to Upgrade the Entire Internet
In 1958 the United States created the Advanced Research Projects Agency to compete with Soviet technology (The USSR launched Sputnik four months earlier). Eleven years later the worlds first packet switched network, ARPANET,was born. Research continued, innovation flourished and on January 1, 1983 the National Science Foundation unveiled the worlds first wide-area network and the TCP/IP protocol. The modern internet was born.
The internet is used today in ways original architects did not imagine twenty years ago. Likewise, the world has changed a great deal since then. Phone lines were replaced by broadband which will be eventually replaced by fiber. Data measured in bytes seems quaint compared to the terabytes (even petabytes) of data regularly exchanged today. Even the original intent of the internet—the exchange of scientific research—has been far eclipsed by its use in commercial and social activities. The fundamental underlying architecture of the TCP/IP network is rapidly reaching its limits. It’s time to reinvent the internet.
The common set of rules for how computers talk to each other over a network is showing signs of age. The architecture originally intended for any computer in the network to be reachable by its network address, is the unique four-part number. The names we type into our browsers today are all connected to some network address (for example, www.google.com is 66.102.7.99). Because of the way network addresses are structured, a decision made over 20 years ago, there can only be 4,294,967,296 unique addresses so Network Address Translators were invented to extend them. Additionally, the meteoric rise of malware and unscrupulous use of the internet gave rise to firewalls. These evolutionary necessities make addressing each individual computer, a fundamental principle of the original design, impossible.
As more and more generations of technology are stacked on top of the existing TCP/IP architecture the system becomes more rigid, inflexible and incompatible. NAT devices are incompatible with many types of security encryption. Almost all forms of internet security are incompatible with transparent web caches. Traceroute and other web diagnostic tools are incompatible with transparent web caches. NAT devices are not compliant with dumb networks. Firewalls make the development of new protocols very complicated. The more layers of code added to the architecture, the harder and harder it becomes to develop new and innovative products. While the internet serves the majority of the world very well today, it is likely that the increasing inflexibility of the system will soon lead to slowed growth, depressed innovation. It’s time to take what we’ve learned over the last 20 years and start building ourselves a network that will support today’s and tomorrow’s needs, evolve gracefully over the long term, and sustain itself for the next generation.
David Clark, the internet’s chief protocol architect and one of the fathers of the internet, is actively leading the National Science Foundation’s research into the development of the next architecture. The new internet is currently in research phase and it will be years before we see the next generation of the internet in common use. We already have an infrastructure to build and test it on: LambdaRail—the nationwide optical network that researchers are using to experiment with new networking technologies and applications. We just need to decide its important enough to devote the resources to it.
Clark writes about the original intent and the future needs of the global network:
The original intent of the internet architecture:
- To connect existing networks
- To continue to work even if some networks were lost
- To accommodate a wire variety of protocols (like http, ftp, etc.)
- To distribute resources around the globe instead of in a small number of centralized locations
- To be a cost-effective means of communication
- To be relatively simple for end-users to connect
New requirements for a next-generation network that are not part of the current design:
- To be mobile, dynamic and flexible
- To be auto-configurable on the end-user side and support per-network policy constraints
- To support dynamic switching of connection method, important for creating consistent connections for mobile devices
- To support dynamic reallocation of system resources, for example better connections for emergency services during a disaster at the expense of common users
- Built-in security encryption support to create a standard and embrace expanded internet commerce
- To provide users with more choices and control over what route their data takes to get from one place to another, much like choosing a local and long-distance telephone service.
- To make data packets trackable to combat malicious internet use like spam, phishing, hacking and bot farming
If the new architecture sufficiently meets the needs of the primary organizations that would fund its development (service providers, governments, universities, large corporations) we may see these changes sooner than we realize. Likewise, if the new architecture promises a spam and virus-free internet, more secure e-commerce and ubiquitous connectivity, the market will undoubtedly demand it. Isn’t it’s time for an upgrade?
Further reading:
Developing a Next-Generation Internet Architecture (PDF essay)
Tussle in Cyberspace: Defining Tomorrow’s Internet (PDF essay)
Tags: IdentityMine
You can comment below, or link to this permanent URL from your own site.
13 September, 2007 at 3:10 pm
Great article, Jason.
25 March, 2008 at 12:48 pm
So is this more than just switching to IPv6?
I know our server admin was bemoaning Ireland’s complete denial of the necessity of switching to IPv6 – but I get the sense that you’re talking about more than just the sheer potential volume of IP addresses that should be available.
25 March, 2008 at 12:48 pm
IPv6, while an improvement over IPv4 (128 bit addresses instead of 32 certainly means more addresses than we can reasonably forsee in the imaginable future) is not truly “rethinking the system”. IPv6 is merely an upgrade to the existing TCP/IP protocol while attempting to maintain as much compatibility with the existing codebase as possible.
IPv6 will not mean the end of spam. It will not mean dynamically self-configuraing mobile networks, It will not mean policy-based network access. It will not mean prioritized bandwidth reallocation. it’s just a way to extend the life of the system we already have. IPv6 is a necessary stopgap to solve our immediate problem of network address exhaustion but it is not a robust enough solution to fix the underlying weaknesses and flaws of the TCP/IP protocol architecture itself.
Sone of the chief evidences of the need for an entirely new architecture are key features built into IPv6. The extensive support for tunneling. Additionally, to maintain compatibility on IPv4 networks, IPv6 needs to encapsulate IPv6 packets in IPv4 with what essentially becomes a link layer. This is a perfect example of “spaghetti code” developing in the system that adds complexity and decreases the potential flexibility of future applications and protocols.
Additionally, the explosion of Network Address Translation devices, one of the evidences that TCP/IP is creaking under the weight of global usage, is slowing the adoption of IPv6 to the point where only government mandate will encourage organizations to switch (the US Federal Government will complete the transition for all their official networks by 2008). Thus far, only Japan, China and Korea have well-thought out plans for embracing IPv6. The rest of the world will live in a state of semi-chaos until the upgrade is complete, which may take a decade (some say more).
IPv6, while necessary, is an incomplete solution to the problems of the network architecture and will certainly complexity and confusion to it. What I (and David Clark) propose is a complete ground-up rethinking of the way the architecture is designed, from a clean slate, using what we’ve learned over the years to build a unified, cohesive, highly extensible architecture that takes advantage of new technology, addresses current problems and is expandable and modular enough to last well into the future. Any programmer will concur that once a product has been added onto enough times, eventually the time comes to refactor it. The global network architecture has reached the point where it’s time to refactor.