Crypto: Parallels to Packet Switching

There is an interesting objection to crypto that goes something like: crypto is all a rip-off because nobody actually needs decentralization which makes things slow and inefficient (“decentralization is a bug, not a feature”). I was somewhat surprised to see this objection from Michael Seemann, a German author, whose criticism of the power of the big platform companies I respect. In a Twitter exchange (in German) he pointed me to something he wrote a couple of years ago as predictions about the recentralization of the internet. Clearly Twitter doesn’t lend itself to differentiated answers, so I am writing this blog post instead.

First, a bit of history. Before today’s packet-switched world we had circuit-switched networks. For quite a long time the telcos, which were operating those networks, insisted that they were more efficient and that their centralized control was necessary to maintain this efficiency and cope with the demands of high bandwidth applications such as voice and video. I am sure that some of the telco people who made that argument genuinely believed in the superiority of their networks but others engaged in propaganda and lobbying solely to maintain power. Incidentally, that hasn’t gone away as can be seen in the ongoing contentious debate around net neutrality, where – to nobody’s actual surprise – it was revealed that the telcos had flooded the comment process.

In his post, Michael Seemann points out that there has been some degree of recentralization of the internet, through companies such as Level 3. One could also add former USV portfolio company Cloudflare to that phenomenon. To take some degree of recentralization at the network layer as evidence that the decentralization project has failed there also is, however, misreading history. While Cloudflare adds some important efficiency to the network, it is only healthy to the degree that theres is meaningful competition. And that is very much the case, with many companies offering competing services, including Amazon, Akamai, Fastly and others. Put differently, yes there has been some degree of recentralization but the power of today’s network providers is negligible compared to the power an AT&T or Deutsche Telekom once held.

Now on to the question of crypto. In the Twitter exchange, I argue that crypto provides a crucial counterbalance to forces of centralization at the application layer. The web gave us permissionless publishing, but as a stateless protocol it required centralized players to maintain databases. As a first approximation the market power of companies such as Amazon, Facebook and Google arises from their ownership and control over large databases. Crypto is a breakthrough innovation because for the first time it gives us permissionless decentralized databases.

It should be obvious that decentralization comes at the expense of some degree of efficiency. But efficiency isn’t the supreme all-overriding concern, we always need to also consider power and resiliency (btw, the same is true in other domains, such as energy). The question then becomes the same as at the network layer: how much inefficiency are we willing to accept in a tradeoff? Here I believe we are with respect to crypto where we were with packet switched networks in the first few decades of those. From the perspective of operating centralized databases (circuit switched networks) these things look like toys. Cute, but useless for “real” work. And of course in a strong historic echo we are getting the operators of the centralized databases (e.g., Visa) asserting their superiority in equal measure because they actually believe it and because they are trying to protect their position of power and profits in the market.

Michael Seemann then goes on to argue that crypto is really the worst of both worlds because (written in 2019) he predicts recentralization and points to a recent post about Bitcoin as confirming his predictions. I will readily agree that we have a long way to go on performance and that recentralization is always a legitimate concern. At the same time in 2021 there are lots of reasons to believe that we are far from the ultimately achievable mix of decentralization and performance. Ethereum is well on its way to Proof of Stake (POS). There is fast and furious innovation in Layer 2 performance for Ethereum with multiple credible projects. There are alternative POS chains, such as Solana and Algorand that can achieve thousands of transactions per second (TPS) at Layer 1, with credible plans to scale far beyond that.

Finally let me come back to what prompted my Twitter reply to Michael Seemann in the first place. He wrote that he supports an outright ban of crypto. To me that is roughly the equivalent of calling for a ban of packet switched networks when they were in their infancy. That would have locked us into a world of communications controlled by the likes of AT&T and Deutsche Telekom. It is also worth considering what it would take to enforce an outright ban on crypto: essentially one would have to ban any and all software that can participate in a consensus protocol. That implies a degree of state control over computation that absolutely horrifies me.