Good idea!
For Eth1x historians, and for context, below is a summary related discussions. I copy/paste quotes here since some channels may not be publicly linkable.
In 2019 and early 2020, we discussed Alexey and Igor’s breakthorugh experiments (1, 2, 3, 4) which made statelessness feasible by using a cache of recent witness data.
On February 15th, I suggested something which resembles the Alaskan Pipeline:
How about a future-cache in the form of a consensus tx pool?
The closest thing that I can find is Tendermint, where txs are propagated before they are committed.
Blocks would include some txs without witnesses, and these txs would go into the consensus tx pool after a delay period during which their witnesses propagate. To accommodate Turing-complete (modulo system limits) contract execution, some witnesses may need to be sent with the block.
On March 3rd M H Swende wisely said that caches bring complexity:
I found the write-up: Caching.md · GitHub (still a bit unpolished).
Example: if a client has a consensus-mandated cache of N blocks, but a new block comes along, where the proof is missing a piece of state that is present in N-1. So unless the client correctly cleaned out every last remnant of N-1, it will incorrectly deem the block+witness as valid.
And if there’s a reorg, so it needs to evaluate a sibling block, then it needs to reconstruct the previous state of the cache (or even several generations back). And in order to have such fine-grained control, we need journalling. And in the end, we’ll likely back this thing with an actual cache that is suited to the actual memory available in the machine we’re running on.
This ended our discussions about caching, and we focused on pure statelessness.
Then, in June 2020, ReGenesis was proposed. ReGenesis uses a form of caching, but is simpler than the others. ReGenesis had potential to get consensus, so we all supported it.
Also around June 2020, there was another idea to decouple witnesses from blocks entirely, see Alexey’s pre-consensus, an avalanche-like consensus for peers to agree on which witnesses are important before blocks are mined.
In January 2021, The Great Alaskan Transaction Pipeline is proposed.