This is a good thread about the MEV War of 2021™️
One thing I will say is that most of the fair methodologies have a downsides themselves:
1. Added latency
2. Lack of guarantees about economic price ordering
3. Extremely unproven in production (similar to ZKPs in 2012)
One thing I will say is that most of the fair methodologies have a downsides themselves:
1. Added latency
2. Lack of guarantees about economic price ordering
3. Extremely unproven in production (similar to ZKPs in 2012)
https://twitter.com/benjaminsimon97/status/1380291321286569986
Why added latency?
Theoretical (@vegaprotocol’s Wendy) and practical protocols (@valardragon) add a >= 1 block commit-reveal from validators OR added rounds of BFT-style message passing. Griefing vectors (DDoS-esque) are abundant + provable models have weak synchrony guarantees
Theoretical (@vegaprotocol’s Wendy) and practical protocols (@valardragon) add a >= 1 block commit-reveal from validators OR added rounds of BFT-style message passing. Griefing vectors (DDoS-esque) are abundant + provable models have weak synchrony guarantees
Recent papers from @algo_class and Joachim Neu show lower bounds on these latencies and it is very unclear if the practical implementations even come close to saturating these bounds (Kelkar, et. al get to weaker bounds in their paper) 
2. MEV Auctions have 1 nice property: they separate the Veblen good-like bidders (liquidators/arbs) from “daily” txns and guarantee price ordering as a function of transaction _value_ is preferred if auction is appropriately ε-close to DSIC (@suryabakshi has more precise conds.)
All of the fair ordering solutions are oblivious to transaction value / content and only care about consistency of a distributed priority queue vs users expressed preferences/DSIC-style guarantees
Is this better? I’m not sure — user intent now relies on latency more than before
Is this better? I’m not sure — user intent now relies on latency more than before
3. Many of the fair order protocols are quite complex with good reason: they’re solving a much more difficult combinatorial problem of optimal ordering of interlacing permutations (one for block ordering and one for txn ordering)
order matters ➡️ can’t do greedy maximal matching
order matters ➡️ can’t do greedy maximal matching
Aside: I once spent an afternoon trying to map the permutation matching in fair ordering protocols to maximal matchings on bipartite Ramanujan graphs (which admit decompositions by interlacing symmetric polynomials), but got stuck and got why Kadison-Singer was *really* hard
Back to the main narrative: this complexity means that we don’t know how well these fair ordering protocols compose with main consensus except in synchronous BFT settings. This is bad for practical Blockchains — not impossible, but will need a helluva lot of engineering
MEV auctions are a lot easier to reason about, especially with regards to how they compose with consensus and even then, one has to be diligent and careful with practical implementations
Summary: fair ordering protocols are a real solution, but they’re far from perfect (both theoretically and in production code). MEV auctions are much less technically difficult and don’t worsen UX as much (which is why I took offense to @AriJuels usage of "theft")
Current fair ordering code reminds me of SNARK code right after the 2012 Gentry paper came out — raw, full of potential, but should be miles away from billions of dollars of assets until compositional guarantees are proven in theory and production
And I will say that I’m excited to see @arbitrum’s code for this – their research + that of @vegaprotocol is definitely the closest to making a convincing argument for fair sequencing chains
• • •
Missing some Tweet in this thread? You can try to
force a refresh
