The tension between centralized and decentralized network design is a common theme in network science. Even a “decentralized” system is actually a continuum, not a fixed state. A system moves along an axis degree of distributedness, and depending on its purpose, there’s not one ideal state. Here Gordon Brander writes about the tendency of networks to evolve, beginning as decentralized ecosystems, evolving through stages of consolidation, then collapsing back apart into discrete connected parts.

Here is a map of the internet. The first thing you might notice is that the internet is not evenly distributed. Instead, we see the emergence of densely connected hubs—centralized islands in the net.

This kind of thing is called a scale-free network. It seems that something like scale-free structure emerges repeatedly within evolving networks, including on the internet, the web, social networks, airline routes,co-authorship in scientific papers, power grids,inter-bank payment networks,Bitcoin mining, train routes, gene regulatory networks,protein interactions, ecological food webs,oligarchies, neural networks…

In fact, it turns out that almost all real-world networks have degree distributions with a tail of high-degree hubs like this. (Newman, 2018. Networks.)

If you see a pattern emerge over and over, it’s a solid bet there are evolutionary attractors pulling the system in that direction. And yeah, it turns out scale-free networks have strange and important structural properties.

Scale-free networks are so common largely because of the efficiency they afford. Fully evenly distributed ones might be “better” in some sense from the perspective of survivability or resilience, but that the cost of highly inefficient function:

Hubs act as routers through which information can travel from A to B to C, efficiently. Scale-free networks areultra-small worlds. Hubs let you jump around the network in just a few hops. Distances in a scale-free network are smaller than the distances in a random network. This is why airlines build hubs.

The argument (which is particularly common in web3 circles) to fight for decentralization over all makes light of the advantages that competing scale-free networks will have in costs and efficiency. If you compare two crypto ecosystems with differing philosophies on this like Ethereum (prioritize decentralization, still care about usability) and Solana (prioritize usability, but still cares about decentralization), the second one has some massive advantages in driving the usage necessary for its existence. Again, one is not objectively better, it depends on its purpose.