Limitations of Blockchain: Why do we need stateless infrastructure?

Author: polynya Source: mirror Translation: Shan Oppa, Golden Finance

I admit that I have talked about strictness many times global consensus mechanism and touted it as the unique feature of blockchain. But I rarely mention alternatives because I studiously avoid discussing infrastructure (infra). The obsession with infrastructure has engulfed the entire cryptocurrency space to the point of eclipsing everything else. However, I realize that people only focus on blockchain infrastructure and rarely talk about stateless infra - maybe because there is no stateless infra to promote. So now I’m going to feel less guilty and talk about stateless infrastructure.

What is stateless infrastructure?

First, let’s define “stateless infrastructure”. Some may say that UTXO, DAG or any chain is stateless. Others may consider a chain with stateless clients to be stateless. Even some solutions that I call stateless have state - albeit in a different form than blockchain. So maybe we need a better term to describe this type of infrastructure, feel free to chat with @polynya on Farcaster for suggestions.

For the purpose of this article, however, we will stick to the following definitions: Blockchain = any chain that achieves real-time strict global consensus (including rollups); Stateless infrastructure = is not a chain and has no consensus (or loose consensus), but a distributed complement that interacts with the blockchain above.

As we move into a hybrid application that uses blockchain only for strict global consensus, and primarily for finance and identity In this new world, we will see almost all computing, data, etc. happening on non-blockchain infrastructure.

But first, we need to change the perspective when discussing stateless infrastructure. In the blockchain world, we are used to the honest majority assumption, so it is very important to run as many nodes as possible. However, outside of blockchain, most things operate on the assumption of an honest minority - as long as there is one honest participant, all is well.

I think we can roughly divide these into three categories:

1. Server - This one is obvious and familiar to everyone.

2. Servers with redundancy - By adding redundancy and allowing anyone to start the server, you can achieve a degree of decentralization while maintaining the maximum efficiency of traditional servers.

3. Stateless Infrastructure - This is a broad category in itself. They are maximally decentralized structures that operate peer-to-peer but without consensus (or loose consensus). So you take the "honest minority assumption" and even a network with only 10-100 nodes is fully decentralized. But it also exists for certain types of stateless infrastructure that only have one node, but through validity proof, they don't need more nodes. (These are called ZK co-processors.) Stateless infrastructure enables many of the features mistakenly attributed to blockchains—decentralization, privacy, auditability, permanence—but without the burden of consensus. By calling it "stateless", we can feel that they have some relationship to the blockchain, but they have no state (PS: they can have a state, but not a real-time strict consensus state like we associate with the blockchain and strict state transition), so no consensus will be reached. Again, this is a very broad category and I'm sure we'll see many different solutions. IPFS, BitTorrent, Farcaster Hubs, ZK co-processors can all be considered forms of stateless infrastructure.

Why do we need stateless infrastructure?

Today, blockchain is extremely inefficient. For the same calculation, its overhead may be tens of thousands of times that of a server. With new technologies like validity proofs and data availability sampling, we can increase efficiency by thousands of times, but even a final ZK rollup using DAS for settlement will likely have 100x higher overall overhead.

Today, "web2" runs on more than 100 million servers. How optimistic we are about decentralized computing varies from person to person. Megachain maximalists are the least optimistic, arguing that 1 chain can handle most computation. The more optimistic people will tell you that 1% of all computing will migrate to decentralized computing, while the most optimistic people will say 10%.

But let’s take 0.01%, which is a bit more optimistic than the megachain maximalists, but still quite pessimistic. This would require the equivalent of 10,000 servers. Add in all the inefficiencies that come with chains, and you might need a million chains.

With ZK technology, it is actually possible to scale to a million chains, with universal synchronization composability and shared security - which is literally a million times better than megachains.

But this is not enough. Let’s take a multiplayer game like Palworld as an example. According to their developers, monthly server costs are $500,000. Running it all on ZK rollup, the ultimate efficiency blockchain solution, would cost millions of dollars, making it financially unfeasible. (As mentioned above, such use cases are completely impossible without ZK technology)

Instead, we must adopt a hybrid approach while remaining decentralized. Multiplayer games can be run largely peer-to-peer, with game clients running on everyone's computer or console, effectively running a massive decentralized network. Some content needs to be coordinated on the server - these can be moved onto a peer-to-peer stateless infrastructure. Things that require verifiability but not consensus can run on the ZK co-processor. Finally, things that require strict global consensus can run on ZK rollup/validiums.

Finally, it's worth noting that it's much easier to parallelize across multiple cores and multiple machines on a stateless infrastructure, which can lead to exponentially higher scaling.

This will still cost much more than $500,000, but by using the most efficient solution for each task while remaining decentralized, we can achieve things that are simply not feasible today, even for future use New use cases that are not even possible with the most advanced ZK form of blockchain.

It’s all about making new applications possible

Today, we are in a very rigid paradigm. It doesn't matter whether it's L1 or L2 - the hardware limitations on a single chain (actually a single server) are extremely limited. We have seen one after another L1 and L2 reach their limits and become congested in a very short period of time, with even the fastest chains only able to process a few thousand transactions per second at most.

By using new technologies like valid proofs, we can start using multiple chains while maintaining security composability, which is already a huge and infinite improvement.

With stateless infrastructure, we can go even further.

By combining all of this, we can begin to implement applications that would not be possible without these technologies.

What about decentralization?

Stateless infrastructure can actually be more decentralized than blockchain. Achieving strict global consensus is an extremely difficult process that requires expensive Sybil-resistant mechanisms and an honest majority assumption. The block production mechanism is a rule of the rich (proof of stake) at best, and a rule of corporations (proof of work) at best. These blockchains mitigate this problem by running on nodes, but they are still based on the honest majority assumption and require thousands of nodes for the consensus formation process to achieve any level of resiliency.

While stateless infrastructure comes in various forms, most work well with only an honest few assumptions - as long as there is one node, everything is fine. As a result, they can achieve greater decentralization with fewer nodes while reducing points of failure and centralization.

In short, if strict global consensus is not required, peer-to-peer stateless infrastructure is both more efficient and more decentralized than blockchain.

Summary

It’s time to stop debating blockchain infrastructure and start building new applications powered by stateless infrastructure.

Further discussion on Farcaster is welcome, @polynya. By the way, Farcaster is a great example of an application that can only be built using stateless infrastructure, which explains why it succeeded while its blockchain-only counterparts like Steem or DeSo failed. More apps like Farcaster, less L1 and L2.