Fadi Barbàra's Brand

A Centralization Framework

One of the most (ab)used insults in the blockchain space is still centralized. This insult was back in vogue last year with the emergence of the so-called L1-alternatives, which more than once have been accused of being centralized. Their answer was similar to “even ethereum is centralized” or “bitcoin is centralized too”.

Beyond marketing, who is right? Is it possible to be right? Actually yes, just structure the problem well and think of “blockchain” as a set of layers. After that, your thinking about the blockchain will improve radically. Not only on the centralization spectrum, but also on the scalability issue and similar parameters.

For simplicity let’s use two academic papers on the matter. The first one by Sai et al which we will refer to as [Sai20] and the second one by Gochhayat et al which we will refer to as [Goc20]. Of course, many others can be used, e.g. the one by Kiayias and Lazos which is specifically focused on blockchain governance.

For example, [Sai20] sees the blockchain as a set of protocols and algorithms belonging to one of those layers:

  1. Application Layer: Dapps, basically
  2. Contract Layer: the smart-contracts
  3. Incentive Layer: Reward issuance, distribution and Punishments
  4. Consensus Layer: the Proof-of-…
  5. Network Layer: Node topology, dissemination, validation
  6. Data Layer: Storage and Cryptographic Primitives

Just by doing that, we have a better way to talk about the centralization of different projects in a meaningful way. For example I can say that Project X is more centralized at the data layer than Project Y (maybe because there some data availability problem or there are very few archival nodes) but Project Y is more centralized at the incentive layer than Project X (since Project Y is a delegated proof of stake where profits heavily go to the more elected validators).

[Got20] has a similar more-coarse layer-splitting:

  1. Governance Layer: the Consensus+Incentive Layers of [Sai20]
  2. Network Layer: Node topology, dissemination, validation
  3. Data Layer: Storage and Cryptographic Primitives

More-coarse doesn’t mean the [Goc20] work is less precise, since sometimes a simpler approach is better than a difficult one which nobody uses (especially on Twitter…). And the example above does fits for this different way of layer-splitting

A side benefit of the “layer view” is the possibility to quantify how much centralization a project has, specifically for each layer. Both the [Sai20] and [Goc20] papers are very good at that: easy to understand equations for each layer are enough to talk about this topic. No need for complex non-rigorous and abstract words.

You now see why the [Sai20] and [Goc20] papers are good for our goal. The authors’ blockchain split into components (layers) and their focus on all the layers is an important framework to keep when talking about seemingly abstract concepts. Specifically: there is a way to make concepts practical and well-defined.

The goal, as always, is to avoid saying meaningless (and marketing-friendly) things, such as “Project X is more centralized than Project Y”, or worse, “Project Z is centralized”.

I hope you liked this post. If you have comments, the best way to express them is by going on Telegram and comment on the dedicated post on the group.

Alternatively you can write me in private on Telegram or an email at me at fadibarbara .it .