Internet Infra - Is the Internet Broken?
Is the Internet Broken Blog - 4 November 2020
In response to the question as to whether the Internet is broken, my view was that despite the core Internet protocols having been designed in the 1980s and earlier, they have worked considerably well and allowed the Internet grow from a relative handful of hosts to billions. The Internet protocols also do not require a hierarchy or central point of control which undoubtedly makes them attractive to deploy, which in turn has contributed to the immense growth of the Internet.
The problems with the Internet are mostly related to security which was not inherently designed into most of the protocols for various reasons.
The Internet was smaller and therefore relatively more trustworthy when the protocols were designed, but there were also practical reasons such as the limited processing capability of hosts and network devices, and possibly a desire to not to overcomplicate the design of a system that was unproven and yet to become successful.
However, the security issues have long been recognised and new protocols or extensions to existing protocols designed to address these - for example, TLS, DNSSEC, DNS-over-TLS, NTS and RPKI. The problem has been the slow deployment of these technologies, in many cases due to lack of awareness and lack of training, but also the need to upgrade or replace equipment in some cases.
This illustrates the problem of attempting to replace the existing Internet with new designs, especially those that are radically new possibly incompatible with the existing Internet. It should be remembered that IPv6 was designed to be the Future Internet back in the mid-1990s and whilst it was not a radical departure from IPv4, did offer several security enhancements. Yet in the two decades since IPv6 started to be deployed and despite depletion of IPv4 resources, just over 30% of the Internet is using IPv6.
Several Future Internet initiatives have been started over the past decade such as RINA, NDN and SCION that promise to address the known problems with the Internet and offer a different approach to accessing network resources. Some of these ideas are actually quite interesting and deserving of further consideration, but are as yet unproven in terms of widespread adoption or scalability, and do not particularly seem to address problems that cannot already be solved on the Internet. So I would see their usefulness more as technologies that can be utilised as virtualised functionality that can be added to the Internet rather than an outright replacement for it. For better or worse, there is simply too much of an incumbent base of network engineers training in Internet technologies, not to mention installed equipment, to transition away from in the foreseeable future.
In response to the second question regarding the postulated establishment of a common transit provider run on a not-for-profit basis that could provision cheaper connectivity between IXPs, this is likely to have limited benefits. Many if not most countries already have or at least had telecoms infrastructure ostensibly built for and operated public benefit by monopoly telecoms providers, and these are often categorised by high prices, poor efficiency and limited consumer choice.
There is arguably a case for public funding and subsidy of new technologies in order to ensure widespread deployment and therefore public benefit, but it becomes more questionable when a technology becomes well-enough established that commercial operators see benefit in entering the sector. Having a single common carrier also leads to arguments over who’s responsible for performance issues, how quickly upgrades are made, and barriers to new operators being able to offer cheaper prices. This dimension becomes even more complicated when crossing national boundaries as envisaged here, how this public Internet provider would be controlled, and even the possibility that it could fall under the control of national governments with less than benign intentions.
This said, this approach may have application in particular sectors such as research and education networking where efforts are underway to coordinate the establishment of a global networking infrastructure that can service the needs of academic institutions and scientific facilities around the world, particularly in the somewhat under-served Southern Hemisphere. And this approach may also be worth considering in landlocked regions of continents such as Central Asia and parts of Africa where there is limited competition and bandwidth is extremely expensive, but here the political dimensions will invariably be more difficult to address than the financial and logistical issues.
I would like to thank the organisers for inviting me to participate on the panel.
About the Author
Kevin Meynell works at the Internet Society as Senior Manager, Technical and Operational Engagement supporting the deployment of key Internet technologies including Routing Security. He previously worked for JANET, the UK NREN, before joining TERENA (now the GÉANT Association) where he worked for the next 16 years on activities including the 6NET and 6DISS IPv6 deployment projects, eduroam, the Global Lambda Interconnect Facility, the TERENA Certificate Service and TF-CSIRT, as well having responsibilities for NREN Development Support in Eastern and Southern Europe, and Central Asia. After leaving TERENA, he worked as the Manager of the Shibboleth Consortium that develops the widely used Shibboleth web single sign-on software, before moving to APNIC as its Head of Training in 2014. He joined the Internet Society in October 2015.