It’s fair to say that most people without a deep understanding of networking probably consider the internet to be a globally decentralized network, which is a reasonable assumption though incorrect. Thousands of internet service providers and telephone companies globally all vying to offer connectivity to householders and end-users give the illusion of a competitive landscape and thus semblance of fair play.
The reality is that the internet has multiple conceptual layers. While the upper layers in P2P networks are decentralized, the lower (physical infrastructure) layers still run on the very much centralized backbone. At the lowest level, there are only 13 major Tier 1 service providers world-wide, with just 4 of them owning over 80% of the global internet infrastructure. Given the right incentives, Tier 1 providers could conceivably collude to shut off traffic for select P2P networks. Then why haven’t they done this by now? The answer is the sheer amount of data that Tier 1 providers handle places a significant barrier to the actual deployment of censorship systems. Still, as we can see with China, Turkey, and Iran, these censorship capabilities are not out of reach.
Using the number of nodes in a system as a metric to gauge its level of centralization is misleading. What actually must be considered is how many different backbone providers and ISPs have a network nodes running on them. Indeed, to be truly decentralized requires that nodes are spread out across as many different countries and with as many different ISP’s over as many T1 providers as possible in order to maximize the number of companies that need to collude to shut down a network.
In the long term, a more competitive internet infrastructure is the solution. Currently, only large players with significant capital can act as backbone providers. Lower barriers to entry, such as with cube-sat delivered internet, leaps in wireless technology, and friendlier government regulations, would allow many smaller players to enter and disrupt the status quo.
Bootstrapping and Discovery
With decentralized p2p networks, there shouldn’t be central server that everyone goes to, no fixed domain or IP addresses. Nodes discover peers by exchanging lists of peers among themselves. New nodes need to kickstart this discovery process by connecting to a predefined list of seeds. In Bitcoin these are defined here and here.
The problem is that these seeds are relied on to pass on other nodes on the broader network, effectively becoming the fixed points that decentralized networks aim to remove. Node operators will wait some time or check with other node operators to ensure connection to the real network to minimize the effects of these central points of failure. To date, there has been no actual node isolation from the network or malicious activity, as presumably we would have heard about it.
Removing the need for seeds is particularly tricky, requiring a radical technological breakthrough, or re-architecture of current systems.
Multicast service discovery is one possible solution, though the network traffic requirements of multicast mean that while it’s possible to use in LANs, it cannot scale to Internet size networks. Given that multicast scales exponentially with network size, it is likely impossible to use multicast on the internet even with many advancements in networking infrastructure technology.
Physically-based discovery systems are another possible solution, allowing discovery without seeds and easy detection of censorship, though they also suffer from scale limitations. Radio spectrum is limited — broadcasters must compete for airtime, and long-range discovery requires significant power. While this may be feasible in large urban areas where low power mesh networks can work, smaller cities and rural areas cannot participate. The usage of physically-based discovery systems at scale requires substantial efficiency gains, but it should be noted that it’s possible to run smaller networks in this way.
The rise of cryptocurrencies has created a large financial incentive for a few bad actors to attempt previously-resource-prohibitive-attacks to be carried out. As these crypto-networks grow, we must increasingly be wary of fringe attacks that were once upon a time not too long ago financially infeasible to attempt. The good news is that these same financial incentives spawned by cryptocurrencies promote research into solving exactly these types of challenges. Only time will tell, however, which side wins the arms race.