Porting this conversation over from Twitter because it deserves more participation/visibility.
Péter Szilágyi (/u/karalabe/): “Can we please stop selling stateless client as a viable solution? Nobody managed to demonstrate feasibility at all over the past 3 years. Proving a single mainnet block updates (reads excl) is 1MB, over an order of magnitude larger than the block. There’s no bandwidth for that.”
If stateless clients are in fact a non-starter based on current bandwidth limitations, this has outsized implications for both the ETH 1.X and ETH 2.0 roadmaps. The former has set aside state rent as a solution to address state bloat (in favor of stateless clients), while the latter relies on stateless clients to handle the challenges of validators operating on randomized shards.
Some background/additional color on these challenges can be found here, in this July post by John Adler and subsequent discussion: https://ethresear.ch/t/open-research-questions-for-phases-0-to-2/5871/2
Key questions to ponder:
– Are stateless clients actually infeasible given the current architecture and limitations? Or are Péter’s concerns perhaps overstated here?
– If so, could a redesigned merkle proof make it work? And if so, what are the next steps and challenges around that task?
– If stateless clients are in fact not viable, what does that mean for ETH 1.X and ETH 2.0? 1.X might have to adopt state rent after all, with breaking-change warts and all. And for 2.0? That would change very fundamental aspects of the planned architecture and require some serious rethinking/respec’ing.
Register at Binance