I had the pleasure of asking Mr Alfour about zk-Rollups and thought others might benefit from the convo so I thought I’d share it here:
Jarrod Woodard:
Hi there! Would you mind giving a introduction?
Jarrod Woodard:
I see that ( in part ) you’re working on layer 2 solutions via side chains and optimistic roll ups. Would those side chain happen to be zkChannels?
Gabriel:
Yup. I worked at NL (even before NL started) for a year, then went on to create ligolang.org, after a while, I wanted to do core dev again (and LIGO Lang’s team became more autonomous, having had a lot of experience). so I went on to Marigold back in November. we started with QoL amendments (that are likely getting merged in H), and are now pivoting to Layer 2s
Jarrod Woodard:
That is awesome, thank you for sharing!
For regular zkroll-ups and optimistic zkroll-ups, they require better hardware than a regular node to process from my understanding. Is that right?
Jarrod Woodard:
In terms of processing the contents of the roll-up. A regular node should be able to include the roll-up in the chain.
Gabriel:
Optimistic Rollups (ORUs) don’t require better hardware, but ZKRUs do
Gabriel:
For the same amount of computation I meant ^
Jarrod Woodard:
Ahhhh, an important distinction, thanks for clearing that up.
Gabriel:
So, if you need to process 1000 tx, ORU will require comparable hardware, but ZkRU will require something a whole lot more (I think x100-x1000 for now, but I’m not confident)
Jarrod Woodard:
Indeed, been serving the community since the beginning.
Jarrod Woodard:
Wow!
Gabriel:
But the main goal of RUs is that you can afford to have higher compute requirements without compromising decentralization
Gabriel:
For instance, let’s say you want to process 100x more tx, with the exact same software
Gabriel:
One way to do this is to say that each node should have 100x more hardware
Gabriel:
But this would hurt security through decentralization: as node costs would skyrockets, way fewer people would be afford to have nodes, and the security of the network ultimately depends on the number of nodes
Gabriel:
However, in the case of RUs, higher computing requirements don’t hurt security, because it doesn’t rely on decentralization
Gabriel:
For ZkRUs, you don’t need any node to be honest
For ORUs, you need a single node to be honest
Gabriel:
Instead of like 2/3rd of the network
Gabriel:
I hope that’s clear enough
Jarrod Woodard:
Thank you so much, that is very helpful. You need the node that creates the roll-up to be honest, correct?
Gabriel:
Nope. You only one node among all ORUs validators to be honest
Gabriel:
Not a particular node, any node, given that any one can validate
Gabriel:
So that’s a very weak requirement: indeed, if no one at all is honest in your network, you have more serious problems 
Jarrod Woodard:
Ahhhh, okay. Other nodes can challenge the honesty and punish them if they are dishonest?
Gabriel:
Exactly
Jarrod Woodard:
Quite true.
Jarrod Woodard:
Nice! …so even a regular node ( hardware power wise ) can challenge the honesty. They don’t need specialized hardware?
Gabriel:
For ORUs, there are two ways to scale:
- more hardware / bigger hardware / specialized hardware, but that’s ok, because you only need one honest validator with enough hardware to verify. you need need everyone to verify
- more rollups. in that case, you only need a single honest validator for each rollup, and each validator only requires regular hardware
Jarrod Woodard:
Would you mind explaining what you mean by ‘more rollups.’
As in more of the transactions on the chain are rollups?
Johann 
:
I think Arthur explained this a bit in his recent video Tezos: Approaches to Scalability. - YouTube
Gabriel:
You can have multiple rollups in parallel, that’s what I meant
Jarrod Woodard:
Thank you so much for the insight Mr Alfour!
