Ellie Davidson ☕️ pfp
Ellie Davidson ☕️
@ellied
X X-posting some stream-of-conscious thoughts on why shared sequencing failed (or hasn't?) https://x.com/cryptobuilder_/status/1885791830752571515 TL;DR: I'm still hopeful that forms of shared sequencing can play a role in L2 composability and decentralization. Shared sequencing didn't work out as a narrative because: it was too early, L2s didn't want to give up sequencing rights, and centralized sequencers offered more benefits to users. Here's how I see things: 🧵
1 reply
0 recast
4 reactions
Ellie Davidson ☕️ pfp
Ellie Davidson ☕️
@ellied
1. Many L2s don't want to give up sequencing rights L2s earn revenue through sequencing, so it makes sense they don't want to give this up. Furthermore, L2s don't want to replicate Ethereum's PBS issues or sequencing model. At the time, many of us thought that Ethereum's PBS issues could be solved through things like ePBS or execution tickets, and that once these issues were solved L2s would be fine adopting Ethereum's model. It turns out this wasn't true. Firstly, we still haven't fixed Ethereum's PBS issues (though we're much closer than we were a year ago!). Secondly, L2s are very opinionated about how their blocks are built. Many L2s specifically want to avoid their users being frontrun or sandwiched, and therefore do not want to hand over block building rights to people like the L1 builders, even if the L2s could keep their sequencing revenue. Other L2s have strong opinions about priority fee auctions, encrypted transactions, etc., which are often at odds with letting others do your sequencing.
1 reply
0 recast
0 reaction
Ellie Davidson ☕️ pfp
Ellie Davidson ☕️
@ellied
2. Centralized sequencing has a lot of benefits Here I'm including the decentralized sequencing narrative because the two were closely linked. Two years ago there was a good amount of public outcry for L2s to decentralize their sequencers. The thing is, though, that L2s were successful because of their centralized sequencers - they offered fast preconfs to users, guaranteed fair transaction ordering, never reorged, etc. It didn't make sense for L2s to give all these good aspects up if they could achieve decentralization other ways (such as through using confirmation layers or different settlement designs).
1 reply
0 recast
0 reaction
Ellie Davidson ☕️ pfp
Ellie Davidson ☕️
@ellied
3. Shared sequencing is just really hard and the narrative was a bit early Shared sequencing has two main components: building blocks for different rollups simultaneously and ensuring safety of cross-chain transactions (such as mints / burns). Both of these are pretty tough, and at the time we still had a lot to learn (and still do!). How do we design shared sequencers that don't require rollups to give up sovereignty, revenue, liveness, user experience, or safety? Two years ago we didn't have satisfying answers to these questions. From the L2s' perspective, they didn't want to jump into shared sequencing if it meant compromising any of those properties.
1 reply
0 recast
0 reaction
Ellie Davidson ☕️ pfp
Ellie Davidson ☕️
@ellied
Since then we've made a lot of progress on these problems. This research wasn't done in the name of "shared sequencing," but it was for the same purpose. I think we're still on the road to something we could call shared sequencing, but the destination looks different than it did two years ago. Here's a few examples of how narratives have changed over the past two years: * L2s need to decentralize their sequencers --> centralized sequencers can be complimented with confirmation layers, censorship resistance, privacy, etc. This way we can retain the good parts of centralized sequencing while still minimizing the bad parts. * PBS with 1 block builder is inevitable --> Let's run builders in TEEs to ensure they follow a certain algorithm, let's build blocks collaboratively (BuilderNet), etc. * Everything should be shared sequenced all of the time --> Shared sequencing is really useful in many use cases, but not in all cases. We still need to solve interop for cases where shared sequencers don't exist.
1 reply
0 recast
0 reaction