Hi Val, thanks for taking the time to review and respond.
I’ll lead off by saying that this proposal is far from mature. I suspect it’s about equivalent to where the rpl.rehab proposal was in March or April.
We acknowledged that our proposal is not as far along, and explained why this should not be a deciding factor. We should do what is right for RP, not pick the option that has been on the table for the longest.
It appears that they then just cross their fingers and hope. This is the core of their whole design.
We went into some detail explaining how we aim to prevent centralisation. If you want to dismiss this as “crossing our fingers and hoping”, that is your prerogative. Anyone can read our proposal for themselves and decide if this is a valid assessment.
It is worth remembering that the rework proposal contains nothing that even attempts to prevent centralisation. Let us be really clear here: The worst-case scenario under delegation (100% of delegation goes to centralised entities, allowing them to launch as many validators as they can afford), is the default state of the rework proposal.
The rework increases centralisation risk in two separate ways:
- Removing the collateral requirement opens the floodgates for centralised entities to take as much of the validator set as they can afford.
- The design of UARS ensures that these centralised validators will profit more than home stakers
1kx assert that pDAO delegation will favor “protocol-aligned NOs” This is not what has happened in real world delegation systems
Firstly, we already acknowledged that delegated staking systems tend towards centralisation (just like every other staking system, delegated or otherwise) because economies of scale exist, and we need to specifically consider how to avoid this.
Secondly, comparing the Solana Foundation (a centralised non-profit foundation) to pDAO (a DAO) does not seem that useful in terms of their respective views on decentralisation, number of participants, diversity of opinion, and so on.
Finally, under our proposal NOs are profitable without pDAO delegation. pDAO is “delegate of last resort”, not the only option to be profitable. Suggesting that pDAO will be required to act as kingmaker in order for NOs to be profitable is misleading.
1kx assert that the rpl.rehab proposal is a bad deal for NOs and a bad deal for RPL holders. At a given size and commission, 100% of revenue flows to those two groups (under both the 1kx and rpl.rehab proposals). It’s not possible for either proposal to be better or worse for both of those parties, since they sum to the same number.
The inputs sum to the same amount. The outputs - from the perspective of an individual NO and the RPL holders - do not, because of the rent-seeking nature of UARS.
Your phrasing here is misleading in two ways. Firstly, you are leaving out a critical distinction between “those two groups”. Consider an ETH-only NO under each proposal.
Under our proposal, those rewards go to the NO running the validators, and the RPL holders who delegated to that NO (which might include pDAO), and maybe pDAO if the node is undercollateralised (used for RPL buys which benefit all RPL holders equally).
Under the rework proposal, those rewards to go the NO running the validators, to ETH+RPL NOs who have sufficient RPL staked, and maybe to providing value to non-NO RPL holders (assuming surplus_share is >0%, and assuming the “replace surplus_share with increased voter_share” option does not pass).
Pretending these are the same two groups is grossly misrepresenting the situation, and takes advantage of the fact that this is a nuanced issue which is easy to misunderstand.
Secondly, focussing on NO rewards, “At a given size and commission” is a very indirect way of saying “if a NO stakes the max amount of RPL”. Saying it bluntly - “NOs only receive max voter_share if they have max RPL collateral” - acknowledges that the rework does not solve the “running minipools is primarily speculative” issue in any meaningful way.
And that is why our proposal is a better deal for both parties: We do not need to siphon off some of the value to pay rent-seekers. Value goes to those participating in the system (the NO and the RPL delegates).
I am unaware of an ethos-acceptable permissionless way to solve this problem and 1kx has not tried to provide any.
There are two questions here: the first is “should we try to protect the protocol against centralisation or just give up and let it happen”. The second question is “how best can we protect the protocol against centralisation”.
We have deliberately not proscribed any single approach for this - it is up to pDAO and it will evolve over time. We need not rely on a single solution, and can innovate here without requiring further protocol changes. We would be happy to contribute to ideation sessions, and would also like to hear suggestions from the community. Personally I would welcome a discussion with you on how to approach this, but based on your response it seems you believe nothing will work and there is no point trying, which is unlikely to lead to a productive outcome.
This fails to acknowledge the value capture component of the rpl.rehab rework. It’s not disagreeing with it, which would be fine. It’s simply pretending that it’s not there.
If you search the posts for “UARS” you will find our references to it, an explanation of why we disagree with it, and some diagrams demonstrating why it is economically flawed and will lead to overpaying larger NOs and underpaying home stakers. We have acknowledged it, and then explained why we do not believe it is adequate replacement.
Furthermore, some of the proposal authors believe surplus_share should be set to 0% at launch. When it comes time to vote on what to do with surplus_share, one of the options is “give it all to voter_share”. There is a non-zero chance that your proposal would end up with surplus_share either left at 0%, or redirected to voter_share. This highlights the perverse incentives present in the rework proposal - RPL holders would be relying on pDAO (i.e. the beneficiaries of voter_share) to vote to reduce their own rewards. The only people who can vote to increase surplus_share are those who would lose out financially by doing so.
I can borrow a bunch of RPL on aave, delegate it to my own node at max no_commission, spin up validators, undelegate it, and repay aave. Zero RPL is staked (thus no value is captured). All the validators are running and gaining max no_commission minus recollateralization_share.
Please see my response to samus below.
When a node operator becomes undercollateralized, some of their rewards are taken away by recollateralization_share. Essentially this brings back a cliff, except it’s now based on a popularity contest rather than personal funds.
It is based on market forces, and/or personal funds depending on preference. I kind of see where you are coming from by calling “market forces” a “popularity contest”, though I am sure we would agree that financial markets have a little bit more nuance to them than popularity contests.
From a game theory perspective, any RPL holder can offer to top them off with a delegation in exchange for 90% of the recollateralization_share. It’s unclear to me why a market for such things (which are punishing to NOs) wouldn’t arise.
Please play out the game theory. You offer 90% of the recollateralization_share, what happens next? Either the NO accepts it, or someone else offers 89% and the NO goes with them instead. But wait! Someone else offers 88%, and someone else offers 87%… this continues until we reach the fair market rate for delegation, i.e. the delegate_share. The market is built into the delegation system.
The flip side is also true! Imagine I’ve delegated 110 RPL to you and that gets you 110% of your minimum. I can threaten to take away 11 RPL (I can just move it to another node),
Again, this is simply market forces at work. You can take away your 11 RPL, then I can either attract another delegate to replace you, or wait for recollateralization, or provide the 1 RPL myself. Given those three options I could calculate the ROI of each and chose the optimal one for me.
This is all just supply/demand economics. We use market forces to make the system more efficient.
When a Node Operator sets no_commission below the maximum allowed, some of that goes to RPL delegators and some goes to rETH. This means that if anyone can instead make an agreement in an outside market,
What you are describing is the “whale marriages” system, where two parties join up to provide the requisite amount RPL and ETH. Your argument here could be rephrased as “people who do not need the delegation system will not use the delegation system”, and yes, that is correct. In a permissionless protocol we cannot prevent this, nor should we want to. We are not going to force people to use delegation just for the sake of it.
The point of delegation is to remove the RPL requirement from the NO and allow others to provide it on their behalf. If the NO does not need anyone to provide RPL (because they have their own, or a whale marriage, or an Aave loan, etc.) then of course they do not need to use delegation. In terms of protocol revenue your scenario is identical to someone who provided their own RPL collateral. The max_commission_rate param ensures these NOs are providing the same baseline rETH APY as every other NO.
Hi @samus, thanks again for your feedback and initial review.
What is missing from the attack vector description is that recollateralization_share is used to purchase RPL and stake if on behalf of pDAO. With that in mind, let us run through this scenario from the perspective of a potential NO with huge ETH resources:
- The NO needs x RPL to deploy y validators. They borrow x RPL from Aave, stake it on their node, and launch y validators.
- The NO can not unstake their RPL for n epochs. They are exposed to price/liquidation risk until they can unstake.
- At the earliest opportunity they unstake the RPL (for sake of argument, assume to zero), and repay the loan (plus interest).
- They are now undercollateralized.
- The validators continue to run, providing revenue for the protocol.
- recollateralization_share is used to purchase RPL from the market (increased buy pressure) and delegate it to the node (reduced market supply) on behalf of pDAO (increased protocol-owned yield).
- This increased buy pressure, increased protocol-owned yield, and reduced market supply, continue until the node is collateralised to the minimal level.
- Once the node is recollateralised, purchases stop. pDAO continues to earn yield on the RPL that was purchased during recollateralization.
So, what is the outcome of this “attack”?
- rETH TVL is increased because the NO was not prevented from scaling up
- Additional buy pressure created due to recollateralization_share
- RPL is taken off of the market, and owned by pDAO
- pDAO RPL yield increases
The NO could achieve a more optimal outcome by simply attracting delegation, and offering (delegation_rate+extra_reth_share)>recollateralization_share. There is no need to go through the Aave rigamarole, it simply costs more in gas and lending costs.
Please also keep in mind that this is a design decision: we could prevent anyone from unstaking below x RPL, but at the cost of making staking less attractive due to longer lockups.