Yeah, how long are miners required to maintain the copies and from what time period since they can leave and rejoin the network at will?

So you can prove that there were something indeed but not what that thing was specifically (like some specific Image) without the non-UTXO-set, right?

Miners can leave and rejoin the network at will. But they will need to retain the UTXO set as long as they intend to be there, and as long as they are going to continue to try to mine valid blocks. This is because they need the UTXO set in order to make certain that the transactions in their block are valid, and that it won't be rejected by the rest of the network.

Since images and other large data chunks aren't generally needed in order to validate the spending of a UTXO, miners don't really have a good incentive to keep it around. But since the transaction containing the image is hashed and put into the Merkle tree, and since the Merkle root is embedded in the block header, anyone who cares about later proving the existence of the data in the chain can just save the Merkle proof for the transaction. Miners aren't required to retain Merkle proofs on spent (or provably unspendable) transactions either, so they will probably just keep the parts of the Merkle tree that include UTXOs. You should look into Merkel trees, section 8 of the whitepaper and try to understand the economic incentives at play for each party involved. Just keep in mind that the only thing a miner needs to care about is verifying that the transactions inside new blocks are valid, nothing much else.

I don't think any miner stores all the transaction data forever, just the hashes/block headers mostly. https://youtu.be/eSuFtCaSu6o

I don’t expect Unbounded Enterprise to store the entire blockchain. I think the ability to put data on chain that miners can easily discard is a major feature. Unbundling storage and time stamping, in a way. Maybe someone else will store the whole thing 🤷‍♂️.

I don't think that we should be counting on options 1 and 2. The reason is that with time, it won't make economic sense for people to store what they don't need for free. In 2020, it might seem like a cool idea for us fellow hobbyists to drop a few bucks on an extra 2TB disk just so we have a copy, but in 10-15 years, we might be filling up one of those 2TB disks (or even a 1PB disk) every few days. And while storage costs do drop with time, I don't think they'll drop off as fast as the rate of new transactions. Don't get me wrong, businesses who can make money might store things, but I don't think the cost will be as trivial as it is today. For now, if somebody is concerned that their data might be lost over time, I would encourage them to try and keep a copy for themself somewhere. Once it starts to make economic sense and once the demand emerges (i.e. once the cost becomes non-trivial for miners), I predict that we'll be able to pay a storage and archival service to keep records—but only the ones people care enough about to pay for, so probably not the entire blockchain after a certain point.

We are headed towards TB blocks and beyond

There’s a good needle and haystack analogy here.

A hash of nothing?

The hash of the transaction used to put it on chain. If the block headers then are matching you have the proof. If you want to get It there will be entities that will store data and provide it as a service.

You have the proof that there were something but if the miners have pruned the data such as an image by not storing it anymore to have room for other stuff, you can't see the data/image anymore that your hash is linked to, right? It's not the duty of miners to store all data forever, right?

If i want to store my data forever, i have to pay for some entity specialized in doing so and how do i know that the data will be stored forever, by having a contract?

Because you pay for it. You will always find someone that want your money for that data.

You know any service provider that would archive my data forever?

I think@jack can make you a proposal at least for your life time, then lets see if one of your descendant want to pay for your data to be still on chain. How many kilobites you have? Years? Lets say 80?

There are many good options for this problem. Storing data in the UTxO set isn’t something where we have adequate price discovery. For that reason, I wouldn’t setting yourself up to rely on that being affordable in the near term. I’m optimistic that this will become a practical option in more and more cases sooner than later.

We are working on affordable solutions for the pre-price discovery/affordability period. In other words, there should be a way for you to pay us to store your data for 80 years soon.