Its a myth that the whole chain is needed. Comes from BTC where you need the chain to recreate all transactions. It has zero practical use. If you would consider the current state of a blockchain to be "wrong" but everyone agreed to move forward anyway you may as well call it "correct" because by definition the majority decides what is right and what not in these systems. So you dont really need to validate the whole chain. All the splits and hard forks are well known and you can only agree or not use that particular chain. The accrual process of validating yourself is moot. Its not like you could find something new by running the exact same code thousands of other ran over that data before you.
Plenty newer DTLs have ditched the chain validation process completely and you can start at any point in time with the ledger state of that time and start going forward to the present. Whether that is years or seconds doesn't really matter. It all depend of whether you have an accentual use case for historical transaction data or not.
This is really important because second generation blockchains have way higher transaction throughput, integrated decentral exchanges (DEX) and other advanced feature that requires way more storage. Some already have reached uncompressed sizes of several TBs for the complete history.
I guess there is not really a definition. Some would say PoS chains is the second generation but its arguably worse than FBA (Federated Byzantine Agreement) and its first implementation in public blockchain came way after the XRPL (FBA) already existed.
There is Stellar which is kinda a re-implementation of the XRPL with many smaller changes but its also using a FBA.
There is Flare (only test net exists at time of writing) its the first Turing-complete Federated Byzantine Agreement blockchain. Kinda the ETH 3.0. but it a new project that just used the EVM (Ethereum Virtual Machine) for code compatibility reasons.
Then there are DAG based "blockchains" but these are more experimental, as in no one really know if they "work" from a mathematical point of view. I would consider them to be the third generation if someone manages to fully implement it and run a public blockchain.
Plenty newer DTLs have ditched the chain validation process completely and you can start at any point in time with the ledger state of that time and start going forward to the present. Whether that is years or seconds doesn't really matter. It all depend of whether you have an accentual use case for historical transaction data or not.
This is really important because second generation blockchains have way higher transaction throughput, integrated decentral exchanges (DEX) and other advanced feature that requires way more storage. Some already have reached uncompressed sizes of several TBs for the complete history.
See technical documentation for the XRPL https://xrpl.org/ledger-history.html