This is a thread on the technicals of Polkadot ($DOT) v. Cosmos ($ATOM) v. Avalanche ($AVAX)

(PS: huge thanks to those that helped me avoid a documentation spiral, especially @_patrickogrady)

First, why am I choosing these three? Well, simple: they are what's relevant right now and -- more importantly -- they share one common "over-arching trait" of having a network of "multiple chains". But the details get incredibly murky, so let's clarify.

Let's start with Polkadot. Polkadot went about trying to solve scaling in blockchains by adopting a "sharding" model. Because Polkadot's consensus protocol cannot scale to many nodes on a single chain, it tries to get around the limitation by having multiple chains.

These multiple chains are called parachains. However, in a sharded system, where there's multiple chains, you need a "coordinator" to ... well ... help coordinate the transactions.

This is where the relay chain comes. It's the primary chain, but it's "primitive" in the sense that it doesn't support anything besides $DOT transfers plus some other operational stuff, like slashing. It doesn't support smart contracts.

So, Polkadot "scales" by having multiple chains (i.e. parachains each with a few validators).

Currently, however, none of this works. The relay chain is live, but it supports nothing relating to actually coordinating parachains! All it supports is $DOT transfers.

Q: "Wait, so what the heck are the other parachains like Edgeware doing?"

Nothing relating to the relay chain! Think of them as literally independent chains. Currently Polkadot doesn't do smart contracts. Maybe it will in the future.

Q: "Ok, so suppose they will in the future, what is the security attached to Polkadot right now?"

Polkadot, in spirit, tries to follow a "shared security model". This effectively boils down to the following: you bid to be a validator and get assigned to a parachain.

In Polkadot right now there are effectively 250 nodes or so. These are the nodes that can become validators of the relay chain, but you have to bid to become a validator. You can't just make it open (like with Avalanche) because you simply can't accept that many validators.

But *eventually* when parachains work, then after bidding high enough to become a validator of the relay chain, you will be placed randomly to become a validator of parachains. This is what "shares" security. This is not too dissimilar from ETH2 model.

Q: "Eventually, when parachains work, what happens if a parachain becomes corrupted because there's not enough nodes?"

Same thing that would happen to any other similarly sharded chain, like on ETH2: you just slash the validators and move on.

However, the funds are lost! The state is irrevocably damaged and the double-spend happened. Entire system rolls back.

Q: "How would you solve this?"

You don't! Only way is to ensure that the fraud proofs are caught faster than it takes for Bob to double-spend on exchanges.

However, the astute reader will immediately realize that all this means is that the time to finality is as long as the minimum time for running a successful arbitrage, which could be very substantial. So in summary, you get neither security, nor low latency, nor high throughput.

Ok, now onto Cosmos. Cosmos is elegant and clean: you have the main Cosmos chain (called the "hub"), which is where all the Cosmos validators are, and you can add other chains (called "zones") using the Cosmos SDK, but the validators are independent.

Different zones can interact with each other through IBC (not live yet), but ultimately the risk is on the users. Don't transact with an insecure zone. The hub is unaffected by this. Also, the Cosmos SDK is very clean, so all in all Cosmos gets a solid gold star.

Of course, while you do get low latencies with Cosmos, there are downsides. The hub doesn't scale to many nodes (it's based on TendermintBFT, also an underpowered consensus protocol), and thus becoming a validator is expensive. It's not truly an "open system".

Finally, Cosmos hub is a bit "primitive" because of its limited functionality. It doesn't support smart contracts. Furthermore, there's no $ATOM lock-in, which dilutes Cosmos security. You don't need $ATOM to launch a Cosmos zone!

Now, finally, let's get to Avalanche.

Avalanche is a network of many networks. All of them are called "subnets", short for "subnetworks".

There's a main network, called the primary subnet. This is a proper, bonafide chain that basically looks like Ethereum: it has smart contracts, transfers, everything. It's also the most secure subnet, and has almost 700 validators. This is where everything happens.

If all you wanted was a super fast (1 second latency), incredibly open, permissionless, and high throughput chain that supports Ethereum smart contracts, you'd be set with just using the primary subnet.

However, if the primary subnet doesn't fit your needs, then Avalanche takes a Cosmos-like approach of letting you launch another subnet. That subnet can have its own validators (just like Cosmos), *but* with a critical difference: you need to be a validator of the primary subnet.

Why this requirement? It forces security to maximally pool into the primary subnet as well. You can't just launch an arbitrary subnet, you need to "borrow" validators from the primary subnet as well.

I see this as the proper design for a system that wants to optimize for decentralization and *permissionless* security while also enabling other permissioned blockchains to interoperate. To be clear: cross-subnet communication is NOT live on Avalanche yet, just like Cosmos.

So, with Avalanche you get the Ethereum security experience, with 1 second finality, low complexity, high throughput, and full-spectrum of utilities you'd expect.

In a few weeks, the Eth/Avalanche bridge will be live, and Avalanche will be the most sophisticated production-ready alternative to Eth.

To the Ethereum die-hards: think of Avalanche as the best L2 solution you can find for your current DeFi needs, even though it's an L1 :)