When I first heard that a world with millions of blockchains is possible, I kind of lost my mind. Blockchains/shards/roll-ups, whatever you call them, there may very well be millions of them. A world with billions of computers sounded impossible in 1960, too.
In order to get there and get there securely, we are going to need interoperability hubs.
Hub & Spoke
Hub & spoke is where all messages between parties pass through a routing hub. It means that you need to secure far fewer connections and is much more tidy to monitor and upgrade.
Hub & Spoke scales linearly, point to point (or pairwise) scales quadratically
At scale, the number of cross-chain communication channels needed is drastically larger in a point to point (pairwise) network than in a Hub & Spoke one.
Where n is the number of blockchains, the number of inter-blockchain connections needed for each model is shown below:
For 4 blockchains, this isn’t an issue, but even once you get to 10 chains, it starts to look pretty unfavorable for point to point.
It quickly becomes obvious that we are going to need hubs for inter-blockchain communication whether we like it or not. Just like we have multi-chain whether we like it or not.
Examples of Hubs
Hubs are used all through society, from airports, to elevators, living rooms, national currencies, power stations, and sub DAOs.
Perks of hubs
Not only are hubs an inevitability, we should look forward to them. They provide specialised solutions to a number of networking problems which shouldn’t be fulfilled by the base blockchains or rollups.
Hubs are routers
Imagine the lady by the telephone switchboard. You can tell her who you want to speak to, and she connects you up with their line. An interoperability hub for blockchains needs to do the same thing. It needs to forward a message, along with some security guarantee.
Translation
Now imagine if she could translate speech between speaker and listener in real time. This is what Axelar does. Translating messages between virtual environments (EVM → CosmWasm) is computationally much cheaper using low-level code at the core layer of an interop protocol, rather than doing translation inside another smart contract virtual machine.
Hubs increase the security of large networks
A common criticism of hubs is that they are single points of failure, and therefore harmful to security. Contrary to many people’s intuitions, in a large system the use of hubs is beneficial to security.
Monitoring (Contagion containment)
Take the Terra collapse as an example. Axelar network is an interoperability hub, connecting many chains. During the meltdown, Axelar passed a single governance vote which froze all messages from the Terra blockchain within 30 minutes.
This secured 12 chains from the compromised Terra network, and from bridging axlUST and axlLUNA between uncompromised networks. With a point to point model, there would potentially have needed to be 72 governance votes to close down all the channels. More likely, in a well designed point to point model, only 12 votes would be needed — one for each chain which hosts bridged Terra assets. In a crisis like the one that brought down Terra, 12 votes is still much much slower and harder to keep track of than 1.
What if a hub is compromised?
The intuitive risk with a Hub model is that when you hack a hub, you can hack all the chains it is connected to. There is more nuance to this point, however. For example, Axelar enables the community to set hub-wide rate limits per asset and per chain, so that a hacker would not be able to mint infinite axlBTC on every chain. A similar protection is basically impossible for a point to point system, or at least it would require stricter limits while in a healthy state in order to achieve the same level of damage control when things go wrong.
The core insight here is that in a point to point system, the L1 blockchains which are being connected are actually the hubs. Most blockchains are not designed to be interoperability hubs, and to use them as such is a mistake. The way forward seems quite clear to me — that we need infrastructure which is designed for interoperability from the ground up.
Upgrades
Even 130 IQs can probably see that upgrades to cross-chain communication channels are much cleaner in a hub and spoke network. Easier to coordinate, govern, implement. This is a security benefit and allows faster iteration.
Liquidity fragmentation
This is a buzzword that is thrown around a lot in multi-chain. In a point to point system, depending on the path that you take, you get different assets with different security characteristics. Most blockchain ecosystems are already seeing this issue.
Imagine needing to individually value and provide liquidity to 20 different wrapped Bitcoins, bridged from 20 different chains. Add in double wrapping or triple wrapping (multi-hop bridging) and you quickly get thousands of different wrapped assets which are all meant to be interchangeable. Tracking the security risks is a nightmare.
A hub model, in comparison, never allows a time traveller to meet themselves. It only allows one bridge-wrapped BTC or USDC on each chain. To send it to another chain, you send back to the hub, then the hub sends to the destination.
Multi-hub (sharded interop)
Once you get to 100 chains, you probably want more than 1 hub. Think of adding hubs as a decentralising vector in a system which is becoming too ordered. They are a tool which can be added or removed to achieve a balance between two dystopias: chaos (warring tribes) and order (imperialism).
In my view, the ideal technology is one which allows the market to decide how many hubs should exist, and what chains are connected to each. This means we need permissionless, decentralised, generalised interoperability technology which supports a hub model.
Axelar is the only real contender for this. Or at least, they have a head start, and a serious advantage in their team composition. Their founder Sergey designed low level software for dynamic routing of internet packets, and then designed and launched Algorand. You couldn’t really ask for a better qualified person to solve this problem.
Interop hubs need to happen
There’s a lot of lyrical about cross-chain security going around. Much of the dialogue is lacking vision of a truly scaled multi-chain world. Up until this point, I believe that the necessity of interop hubs has been neglected. They are going to be a core part of this ecosystem and should be considered from the get go. Axelar are tackling this head on, and have built a sound blueprint for scaling this internet of blockchains.
Next, I’ll be reviewing a paper on overlay networks for the internet. I’ll liken the scaling issues of blockchains to the internet infrastructure and sow seeds of hope that we can scale this thing.
