I have been an Ethereum watcher and dapp developer for years, and have admired all the EIPs that have come through. However, this EIP is deeply troubling. I think this would be an extremely negative EIP to implement. Here’s why:
Ethereum was touted as the system to build “unstoppable” apps over the years, I loved it. With this EIP, these “unstoppable” apps will simply, well, stop (at least, their UXs/UIs will). It forces substantial and necessary adoption of some a.n.other (unknown and uncertain) protocol entirely outside of the Ethereum system. Pulling the “promise” of data persistence on old apps will be disastrous for the long term reputation of dapp development on Ethereum.
For those who say Ethereum was never meant to store data permanently, that is simply not true. It was! It’s specced that way and therefore used that way. And with this EIP, it will no longer. This is a truly fundamental change to (and destruction of) the Ethereum value proposition. Providing canonical transaction history tightly coupled with canonical transaction generation is CRITICAL to Ethereum’s value proposition. Offloading this entirely outside of Etheruem’s control gives away (and destroys) the future utility of Ether. Why? Because the whole point of canonical transaction generation is that you also have canonical transaction history.
With EIP 4444, it is possible to lose entire chunks of past transaction history. Forever. As in gone. No one knows who sent (or did) what to who or when.
There’s a reason why JP Morgan, HSBC, etc, any of these long storied banks are still around. It’s because you can rely on them having, somewhere inside their big walled offices, a transaction history going back over 100 years. This builds TRUST (yes, centralized trust). But that’s why (other) people come back to them (even if blockchain ppl don’t). The old history may be hand written in log books, sure not convenient, but it’s there.
Now imagine Ethereum was just such an organisation. You go to Ethereum in 12 years time and you ask (in code), what transaction happened on this account 10 years ago? The reply: oh go to Graph/Bittorrent/IPFS/a.n. other, we don’t keep that. You try numerous of these organisations, by some stroke of bad luck, they messed up on your particular EOA/contract (or their tech dies), and it’s gone. Would you trust the Ethereum system, simply because it moved to cool stateless consensus, and therefore decided they didn’t need to include anything as boring as past history anymore? I wouldn’t.
What this EIP fails to realise is that the value of canonical decentralised transactions in almost all real world use cases isn’t canonical decentralised transactions, but canonical history of decentralised transactions. The blockchain that does this will win. Ethereum does both today, but with this EIP, it won’t any more. That’s what I would call the broken promise of Ethereum if this EIP happens, and without a simultaneous EIP that ensures canonical transaction history at the Ethereum protocol level. Perhaps some compromise can be made on time horizon. “Forever” is not enforceable, but say 20 years (for example) is, and good enough for most real world use cases (but even then not good enough for academic records, for example). Remember, the well known banks most of us also use can keep (centralised) canonical records for over a century, and universities can keep records for multiple centuries.