By: Leslie Daigle
Date: September 7, 2009
These days, there are a lot of activities in the world that are focused on understanding and/or developing “the future Internet.”? On one end of the spectrum are national or regional research programmes whose goals are to develop networking technologies that are free of the perceived failings of the current Internet (such as security issues, congestion, and traffic management issues); on the other end are formal specification activities within international organizations that target mainly some hypothetical other network, without regard for how that network might be deployed. However, there’s a paradox: if you can predict the future of the Internet, it’s no longer the Internet.
That rather flip statement embodies a lot of what many IETF participants understand implicitly: that the future of the Internet is not planned; it evolves based on the networks that get built and how they are used (applications and services). The global interdependencies that are inherent within and between networks and their uses make interoperability the best basis for technology development. While these concepts may be well understood by IETF participants, it has become increasingly important to explicitly articulate them and to ensure that they get more generally understood-if we still consider the principle of evolution based on interoperability and innovation the right basis for development of the Internet.
In that model of development, there are three key types of activity from which feedback needs to be fostered: (1) development of technical specifications, (2) deployment, and (3) answering open questions (research). The IETF focuses primarily on the first activity, but it has a close relationship with the third in the shape of the Internet Research Task Force (IRTF). The IETF process itself is deliberately open to and seeking feedback from the second activity: deployment experiences.
The Internet was built on the premise of interoperability based on independent implementations of common specifications: Internet specifications. By focusing on interoperability for passing traffic between networks, Internet standards describe the protocols on the wire without prescribing device characteristics, business models, or content.
The value of this building-block approach is seen in the range and depth of innovation and development in Internet technologies and services. New components-whether networks, services, or software-work seamlessly with existing deployments, as long as all of the pieces correctly implement applicable standards on the network. This makes the field of possible innovations virtually limitless.
Apart from the focus on wire protocols for interoperability, one might say that successful Internet standards share certain characteristics, as follows:
- Freely accessible specifications: All of the relevant written specifications required to implement the standard are available without fee or requirement of other contractual agreement such as a nondisclosure agreement or license.
- Unencumbered: It is possible to implement and deploy technology based on the standard without undue licensing fees or restrictions.
- Open development: In order to have relevance in the resulting standard, it is critical that all parties working with impacted technologies be able to participate in and learn from the history of the development of an Internet standard.
- Always evolving: As the Internet itself continues to evolve, new needs for interoperability get identified. Therefore, the standards that support the Internet must evolve to address identified technical requirements.
Again, these characteristics may be familiar to IETF participants, but they are important to articulate and share.
Deployment Realities: Awareness and Feedback
For newly developed building blocks to work seamlessly with existing deployments, they have to be placed based on some level of awareness of actual deployment realities. It’s not enough to posit a desirable outcome; feedback from past successes and failures, deployment conditions, and expectations of uptake are required throughout the development of new specifications. For example, the IETF encourages this through open participation by all engineers with relevant expertise, as well as the formation of working groups dedicated to operational aspects, such as v6ops and dnsop.
Of course, while the technical specification process views deployment realities as input, broader deployment discussions are important in the identification of critical needs too. That is, operational experience with network usage, new or updated protocols, best practices, and so on are things that are best articulated in groups of deployment experts: people with operational expertise. This is where regional operator group meetings, such as such as NANOG, RIPE, and APRICOT, are key for network operations activities. More-regional and more-focused network operator groups can draw experts to discuss local issues as well as global issues in context. It’s especially valuable to get cross-pollination between these activities and technical specification activities.
Sometimes it’s important to bring back the deployment reality issues to the IETF in a broader context than specific work in a particular working group. This is often the driver behind the Internet Architecture Board’s technical plenary topic selections. The session on network neutrality at IETF 75 (see Plenary, page 4) provided just such an opportunity; it was a chance to hear the perspectives of decision makers (governments and regulators) that are outside the traditional operational network realm. It’s also a motivation behind the Internet Society’s recent media briefing panels, such as the Securing the DNS panel (see page 12).
Looking to the Future: Research
When it comes to gathering data, examining issues, and seeking answers without the restrictions of established environments, organized research is key. The IRTF’s work with the IETF can serve as an important bridge between the world of research activities and the realm of technical specification. This was especially well illustrated in the case of the Host Identity Protocol, which has had concurrent research and working groups examining various aspects of development and specification.
As noted earlier, there are a number of clean-slate research programmes under way around the world, many of which focus on considering known issues-such as security, congestion control, and routing-within new network developments completely independently of the deployed Internet. The research will yield interesting answers to the important how and what-if questions. The next question is, How will the world make use of the answers? It could be through the blanket deployment-from scratch-of the new networks that those research activities propose from those clean slates. However, that could not happen overnight. Alternatively, the lessons learned through those research activities may well inform current Internet building-block developments, because strong evidence of the value of a different direction provides impetus to get development and deployment over hurdles that might otherwise have seemed insurmountable. That means that research activities must be discussed and shared equally within the processes for technical specification and deployment feedback.
Fifteen years ago, the percent of researchers among active IETF participants was higher than it is today. Perhaps that’s not surprising, given that the core Internet then still featured a large number of research networks and nodes operated by academic and research institutions. Nevertheless, there are still researchers who get involved in IETF activities, as witnessed by the level of attendance at an Internet Society cross-regional (Europe, North America, and Asia) future Internet researcher luncheon, described here. In a discussion of the challenges to future Internet research activities, it became clear that one of the significant challenges involves getting a coherent research agenda that is useful for framing funded research activities across regions. That’s one way the specification and deployment activities of the Internet could feed back into the research world. Likewise, highlighting the most promising research results from around the world to the operational and standardization communities will help close the loop on the cycle of activities that constitute this model of Internet development.
For those of us who still consider the principle of evolution based on interoperability and innovation as the right basis for development of the Internet, it is important to recognize, foster, and share the value of interaction between specification, deployment, and research. There is no master plan for the development of the Internet, and so far, that has been a feature. Going forward, the health of the Internet is going to continue to depend on the health of this ecosystem of development.