This is exactly what state expiry will get us eventually.
I don’t think we disagree. I think the only problem we have here is that vast majority of all interesting state is something that will be HUGE if Ethereum becomes the ground for all economy-related activities in the future.
L2s will help there too ofc.
Nevertheless, the amount of interesting state tends to infinity when time grows enough. And that is something that no measure (neither state expiry can solve).
Thus, the idea from Vitalik makes a lot of sense in combination with these types of nodes loke VOPS or Ress.
Users can follow the tip of the chain trustlessly, participate in consensus duties and keep healthy mempools for FOCIL or tx broadcasting.
On the other hand, Dapps and Wallets can keep the piece of protocol-related state only instead of a bunch of useless data they’ll never use for anything.
Then, these Wallets and Dapps can have a much easier time serving requests via RPC to their users specifically which can be verified against trustlessly-asserted roots.
Otherwise, we always need a GOOD_SAMARITAN who will serve this data to users that don’t want to store anything locally. Maybe not now. Maybe not in 10 years even.
But eventually, the fact that this happens means ethereum succeeded and became so big and important that we need to rely on these solutions.
The only caveat to all these discussions, is that this is only possible if the block producer HAS ALL THE STATE. Otherwise, strong_statelessness-like solutions will significantly harm UX unless a ton of effort is put. And won’t enable gas_scaling nor following the tip of the chain at such a low cost.