Having considered this further, I think the situation is more nuanced. I think there are fundamental differences between canister controlled neurons and how they’re implement. I’d love to hear your thoughts @skilesare.
Take liquid staking solutions for example, WaterNeuron’s solution and StakeGeek’s solution were worlds apart. I’m still trying to get a feel for Meta Pool (but the closed source nature of it doesn’t fill me with hope).
I think if the canister in question is verifiably controlled by a sufficiently healthy DAO, meeting a set of necessary criteria, then I don’t see why VP restriction on the NNS side would be needed, or even fair (after all the NNS is a DAO, which also needs to be healthy and is expected to meet the same sorts of criteria).
This may not be an exhaustive list, and would be cool to flesh it out based on community discussion, but I think the two most important things are:
- A sufficiently high Nakamoto coefficient (this is obvious). Large enough to account for reasonable margins of error, with a sufficiently large portion of the VP doxed.
- Non-transferable neurons (which govern the DAO). This should also be obvious. It ensures voters are tied to the outcomes of their votes. Transferability simply destroys this guarantee. Even if it’s possible to subvert transferability restrictions (e.g. IdGeek), it’s still important for those restrictions to be there (locks on doors don’t always keep burglars out, but doors are still designed with locks, and that’s good and proper).
Are these the only two foundational requirements for a DAO we can have confidence in, or are there others?