Mustafa said “However, there could also be some MEV on the actual data layer itself if the rollup is constructed to have a decentralized set of sequencers, allowing the rollup to order the blocks within the data layer.” Can someone clarify what he mean by this?
What is the difference between the “data layer” and the other MEV within each rollup, in this definition?
The basic idea is that you can derive ordering from the DA-Layer. Let’s say we have 3 Rollup Blocks in Celestia. And 123 will result in a different blockchain than 231. This way, we can introduce MEV on the DA-Layer by reordering blocks. The Celestia Validator has the ability to do this. It always depends on how the Rollup interprets the bytes on the DA-Layer.
If the ordering of Blocks on Celestia does not matter (per Rules of the Rollup), then you have the MEV on the Sequencer Layer. TxaTxbTxc will result in a different block than TxbTxcTxa. The Rollup Seqeucner has now the ability for that reordering.
Shared Sequencers are a different concept and, depending on the implementation of the sequencer and the rollup, will have more or less extractable MEV. This is independent of the DA-Layer, so it does not matter if it is on Celestia or Avail.
When the DA-Layer “acts” as a shared sequencer for based rollups, for example, it depends on the DA-Layers implementation and the Rollups’ interpretation. Read here on Based Rollups and an implementation Idea of them here.
In short: Any Based Rollup will leak MEV to any DA-Layer if the DA-Layer can capture anything from reordering.
But I know Celestia can work as a sequencer for rollups. This is why I find some similarities with shared sequencers.
Also, an quote from that podcast:
Anna Rose: And you did say that like you’re actually working on kind of ordering, does Celestia then act a little bit like the sequencer of rollups?
@musalbas : It can do in the sense that Celestia definitely orders the rollups blocks, but it does not mean that rollups can’t also have their own sequencer.