Blog, Notes, and Other Things

19 Feb 2022

Decentralization Is a Quality

Decentralization and distributed system are relatively old terms in computing that are getting more widely used outside of these circles. Most of this adoption can be attributed to one phenomenon, crypto.

Decentralization is as the name suggests refers to systems without a center. This is in contrast to a centralized system where the center is the source of truth, control, dispute resolution, etc. all of which still need to be solved in a decentralized system but are done so without a center. Distributed on the other hand refers to systems that are spread out over multiple things (nodes, processors, etc.). For this post, I’ll be focusing on decentralized systems, but often decentralized systems rely on distributed state1, so I’ll be referring to systems that arise due to both of these qualities interworking.

Decentralization is a quality of a system. To a large extent, it’s also a non-user-facing quality. There is some argument to be made that decentralization will be visible to the user as resistance to single points of failure (and things like censorship etc.) but for the most part, users don’t care if the system is decentralized or not as long as it solves a specific problem for them.

As with all qualities, decentralization doesn’t solve any problem on its own. Rather it needs to be paired with a job to unlock its value. This doesn’t mean decentralization is not useful.

Why decentralize? 🔗

One of the main reasons why the qualities of decentralization and distributed systems are valuable is they make a system more fault-tolerant and guard against single points of failure. This is well illustrated by the Internet, the largest distributed system built by humans. Decentralized systems are also by their very nature resistant to censorship up to a large threshold.

Similar to all qualities decentralization comes with its tradeoffs. For one, decentralization, in general, increases the cost of coming to a consensus, and this is important since consensus is a core requirement for any type of coordination or a job to be done. Decentralized systems by their very nature also make it hard to fall back to an authority for recourse.

In general, the problem space that is tackled by decentralized systems is a (strict?) subset of problem space that is addressable by a centralized system often much more cheaply (at least in the short term)2. But decentralization is an extremely valuable quality in places where there is a lack of trust in any centralized entity. Because of this, it is easy to lament that decentralized systems do not solve any new problem, this might be strictly true but decentralized systems allow to rebuild existing systems by simplifying them by taking out trust on a central entity.

The extent to which this is valuable depends on the extent to which society doesn’t trust the centralized entity being removed, but in a long enough time frame simpler systems with lesser components outlive more complex systems which have an interplay of various components.

At the same time, systems that are good enough with a large number of inefficiencies have a very large half-life especially if there are intertwined with the functioning of society. The current solutions of centralized banks and government is a relatively stable position at least as of now. So the challenge of crypto adoption is to take us from this state to another equilibrium state while removing trust in centralized systems without a lot of changes to the actual job of money. This displacement of the trust in centralized entities will inevitably lead to a lot of resistance as we are already seeing from the incumbents which have this quality as core to their business model and rightly understand changes to it would disrupt them.

Decentralization and Platform 🔗

Looking at the portfolio of crypto-focused VC funds, it’s hard not to notice the fact that most of the startups have the words “platform” or “infrastructure” on their landing page3.

One way to think about the prevalence of platforms in the crypto space is to look at the composites of crypto. Crypto in essence is decentralization + store and exchange medium of value i.e, money. As such crypto companies broadly fall into two categories, those that enable and facilitate decentralizations (for example by building better decentralized protocols, building development tooling for DApps, wallets that allow swapping tokens, DEXs, etc.), and those that enable and facilitate traditional forms of finance such as lending, staking, etc. in the decentralized ecosystem4.

The companies that focus on decentralization are focusing on a quality and as discussed in the last post focusing purely on a quality leads to a platform. And the companies that are focusing on traditional forms of finance are simply building digital analogs of physical systems like banks, exchanges. It is quite possible that once these are built and everything becomes digital and interactable via contracts there might be new emergent phenomena that might not have been possible via their physical counterparts, but for now only the most obvious infrastructure layers of finance are being built out.

One more possible reason for the prevalence of platforms in crypto could be that startups are carrying over the learning from the traditional web2 startups which benefitted heavily from building platforms and extracting platform fees i.e, following the idea of selling shovels rather than digging for gold.

Platforms before Product? 🔗

Historically infrastructure was built with some use case in mind. Though no doubt the infrastructure builders couldn’t conceive of all the use cases that would be unlocked by their effort5, they nonetheless built the infrastructure with at least a use case in mind. Often the infrastructure builders were the first best customers of their infrastructure and they built it mainly for themselves. For example, the first communication cables across the ocean were only introduced after the telegraph was proven as a useful means of communication. In each of those cases, some rudimentary version of a product came before the platform.

But this seems less so in the case of platforms that are built for decentralization where everyone seems to be building a platform rather than a product.

Though platforms provide abstractions for newer products to be built, building platforms without sampling a large enough use case leads to platforms that fail to generalize essentialities that are common for different products6. As such if there is a “revolutionary” new use case that crypto unlocks then some of these platforms may become less useful due to building at a wrong abstraction.

[P.S: I wanted to write this post partly as a follow-up and test case for the previous idea on qualities and business models, but while writing I realized there are so many things I still needed to explore in this space. As such I want to use this post as a snapshot of my current limited thinking in this space and use it to compare it when I have a bit more experience in this field. I will admit even though I have dabbled and followed up with the concepts in crypto for a very long time there might have been plenty of things I might have missed. So I welcome any criticism and corrections on my thinking on this post!]

  1. For a more detailed distinction between decentralization and distributed, see here, or here or more classically the Wikipedia definitions of decentralized system and distributed computing ↩︎

  2. I couldn’t think of a problem that can be only solved by a decentralized system as long as there is a trusted centralized entity. See also ↩︎

  3. See for example the a16z crypto portfolio↩︎

  4. There are also other companies such as consulting firms that build these apps for other companies, art houses that produce NFTs, etc. but all of these are players in the same ecosystem and they facilitate and enable either decentralization or exchange of value. ↩︎

  5. Existing infrastructure not just enables but also constrains future products. Although this is less so in digital products than physical the effect still exists. ↩︎

  6. It could also be possible that platforms that are just digital analogs of staking, lending, etc. do not need to sample a large use case, as this is already done by their physical counterparts of banks and if they can just replicate it digitally with the decentralized quality that would be sufficient. ↩︎