The Internet is Broken: Idealistic Ideas for Building a NEWGNU Network

The Internet is broken, by design. Recent revelations about the abuses by the NSA or GHCQ rarely contain stunning details about new magical technical capabilities of these agencies, but instead merely detail that they have both the budget and the moral framework to exploit known vulnerabilities at scale. The problems of today’s Internet start at the Ethernet layer, where sender’s MAC-48 addresses can be faked, and packets can be intercepted and modified by switches. The TCP/IP layer has the same problems, with routers learning source and destination of all communications, as well as details about the payload (such as port numbers). Routers can also interfere with connections, for example by injecting RST packets. TCP also operates with the assumption that other traffic will be “TCP-friendly”, which is like having a speed limit on roads without enforcement. TLS, the workhorse for today’s “Internet security”, provides “security” only if all of hundreds of certificate authorities operate correctly (which they usually do not), and it comes with a large set of supported cipher suites, most of which are known to be insecure. All of the above facts are well-known and even discussed in ordinary news venues before Edward Snowden decided to expose some of the systemic abuses supported by these design flaws. We will present ideas for fixing the network, with the goal of building a GNU network, where users have the freedom (0) to securely access information (“run” the network), (1) the freedom to study all aspects of the network’s operation (“access the code”), (2) the freedom to distribute information (“copy”), as well as (3) the freedom to deploy new applications (“modify”). Today, monitoring infrastructure, proprietary implementations, traffic shapers and firewalls restrict all of these essential freedoms to some degree.