thanks for the super thorough article @potuz !
i too love the idea of enshrining PBS, but I think the property that I haven’t been able to get around is 3.5.
we have discussed bypassability at length, but in my eyes, this is the property that no design has fulfilled.
additionally, i would like to call out a few points that Barnabé has made at different times. the first is that ePBS should not be constrained to just the “full block auction”. he explains it better than i can in this post: Unbundling PBS: Towards protocol-enforced proposer commitments (PEPC), but the TLDR is that enshrining this specific version of the auction that mev-boost implements (the full block being sold by the validator) is itself and opinionated design decision.
secondly, even the allocation mechanism of giving each proposer a slot may be re-evaluated. execution tickets, Execution Tickets, try to explore this world, where we really wouldn’t benefit much from ePBS, because the proposer themself wouldn’t even be responsible for producing an execution block at all!
lastly, when considering the endgame design of the consensus mechanism, i think also exploring concurrent block proposer constructions is very important! ROP-9, Notion – The all-in-one workspace for your notes, tasks, wikis, and databases., covers the high-level topic and has a link to related documents.
these 3 ideas significantly shape my thinking of the extent of the design space!