What are the cryptoeconomics suggested to solve the DA problem in this protocol?

[Nuhvi]

The same problem @peterktodd noted. There can be cryptoeconomic means of mitigating the problem - the same way cryptoeconomics of Nakamoto consensus helped to solve byzantine generals problem.
We will elaborate much more in bitcoin-dev mail list once Peter publish his review.

• https://xcancel.com/lnp_bp/status/1664973049295257605

Where can I read further about the cryptoeconomics solution suggested here

[Nuhvi]

I thought about this, and I figured, the best any blockchain can do is incentivize miners to keep data, but they can't force them. For example even Bitcoin miners don't necessarily have to store the UTXO set, they can delegate this to a trusted indexer. But these indexers because there is a demand for that historical data, including for validation.

A CSV L1 doesn't have this incentive, but it can introduce it, it can demand that headers commit to either:

1. Full set of previously published nullifiers.
2. Or uniform random sample of previously published nullifiers.
3. Or recency biased random sample of previously published nullifiers.

Random samples depend on the header's nonce.

The first two options requires monotonically increasing storage, which is worse than Bitcoin Utxo set, while the later option requires a bounded set, giving users plenty of time to recover their funds.

I think this, combined with the savings on the cost of validations, might make mining more decentralized, not less.

The caveat here is; this doesn't actually work for the infinite scaling, suggested here, because all nullifiers need to be broadcasted through the p2p network of miners and validators, which reintroduces the scarcity of transactions per block.

Due to the ephemeral nature, absence of validation and no need in knowing PMT content for mining the next header, PMT size doesn't affect system scalability. Thus the PMT may be of (multiple) gigabyte size, committing to billions and billions of transactions.

This has never been true though, because miners can't mine the next header without knowing that the previous one has been broadcasted in full to a vast majority of users, because what is the purpose of mining an L1 where all users have just lost their money? There won't be fees.

It seems to me, CSV is a good idea, but there is so much TPS acceptable before mining centralization becomes guaranteed. So maybe something like tree of chains (sharding) is the only way you can get unbounded TPS.