Florian Suri-Payer

CS PhD @CornellCIS
Distributed Systems, Blockchains, Databases

[Florian Suri-Payer @floriansuri]

@InfernoSOX @NftBearDad You may find our original Autobahn paper a good reference https://t.co/XOPoAADP2i

One underappreciated aspect (latency and throughput aside) is also the novel robustness property ("seamlessness") it provides for real networks.

[Florian Suri-Payer @floriansuri]

@NftBearDad @SeiNetwork @BuildwithSei Love seeing the adoption and spotlight of our Autobahn work -- but would appreciate also candid referencing!

https://t.co/XOPoAADhcK

[Florian Suri-Payer @floriansuri]

@SeiNetwork Awesome to see more people adopt our Autobahn protocol to production.

Great job!

[Florian Suri-Payer @floriansuri]

@Recently_Rugged Love hearing that!

[Florian Suri-Payer @floriansuri]

@richoffcontract @SeiNetwork You can read our Autobahn paper here: https://t.co/XOPoAADhcK

[Florian Suri-Payer @floriansuri]

@0xHeroe_ @SeiNetwork You can read our Autobahn paper here: https://t.co/XOPoAADhcK

[Florian Suri-Payer @floriansuri]

@SeiNetwork Awesome!

[Florian Suri-Payer @floriansuri]

@_patrickogrady @ittaia @siobhcroo That's awesome to hear :)

[Florian Suri-Payer @floriansuri]

@samlafer @ittaia However, if a byz leader **does** propose, then it must respect whatever lane coverage policy you’ve specified.

[Florian Suri-Payer @floriansuri]

@samlafer @ittaia All blocks proposed by correct/honest replicas are guaranteed to be included as soon as a correct/honest replicas becomes the consensus leader. I would consider that "timely".

A byzantine leader (there might be up to f) may always propose nothing -- by pretending to be slow.

[Florian Suri-Payer @floriansuri]

@amitpr @ittaia Importantly, our gracious case fast path does not come at any cost of worse performance of the "slow" path (still faster than DAGs!).
It's simply a bonus that optimizes latency when possible, and is a well-understood optimization applied in BFT literature (Zyzzyva, SBFT).

[Florian Suri-Payer @floriansuri]

@amitpr @ittaia to have n=3f+2c+1 to tolerate some number of crashes in the fast path. Some systems to this -- but we decided not to for simplicity in this paper.

In practice, it is often reasonable to assume that the system is fault-free most of the time.

[Florian Suri-Payer @floriansuri]

@samlafer @ittaia The ordering **within** blocks itself can use orthogonal mechanisms (e.g. the paper I referenced).

Whose blocks, and how many are to be included (per consensus) is configurable via our "Lane Coverage" predicate. S 5.2.3

[Florian Suri-Payer @floriansuri]

@samlafer @ittaia Chain Equality, instead, says that all honest builders have the **opportunity** to contribute an equal amount -- if they put in the work. We think of "fairness" as the outcome being proportional to invested resources. This is the fairness notion that Motorway tries to realize.

[Florian Suri-Payer @floriansuri]

@samlafer @ittaia I encourage you to read the full paper!

[Florian Suri-Payer @floriansuri]

@samlafer @ittaia In fact, in Motorway all 3f+1 builders get their blocks included; this is stronger than the 2f+1 in DAG protocols. We don’t need to worry about orphans.