Thread by @pseudotheos

What is data availability? Why is it important for blockchains? A 🧵 breaking it down 👇

Data availability (DA), much like ZK proofs, is an actively researched area. A lot of problems still need to be solved before implementation, and there are several potential solutions. I'll break down one solution: high dimension erasure code and Kate commitments.

It all stems from one question: how can nodes in a network verify that all of the data from a new block is available, and nothing is hidden or censored?

If the data isn't 100% available, a block producer can smuggle in malicious transactions or impact security.

How do we fix this?

Similar to torrents, the key is redundant data. Using a lot of fancy math, we can create use erasure coding to create fraud proofs, which prove to the node that the data is complete.

Here's an example of a block with DA implemented:

This has some issues still, including the fact light nodes can't currently handle the additional data, it increases computational overhead, creates state bloat, and it requires at least 1 honest node in the network for it to work properly.

This amplifies another problem: Storing more data means the chain gets larger, which is why we need state expiry and statelessness! If we want true scaling, we can't have a single transaction exist as data in perpetuity, as the growth would be unsustainable.

In my next thread, we'll go over statelessness and why it's important (and required) for scalable data availability.

If you want to dive even deeper on polynomial commitments as a potential DA solution:

hackmd.io/@tompocock/Hk2…

dankradfeist.de/ethereum/2020/…

For further reading: arxiv.org/abs/1809.09044

(thanks @epolynya for the link!)