Hey! While I always enjoy our discussions and especially diving deep together to analyze technical and economic system nuances, and all the neat things a DiVa is doing in this regard, unfortunately I think this response is exaggerated and not internally consistent.
You mention systemic risks and irreversible consequences but I see neither the presuppositions for such an argument nor the rationale through which this conclusion is reached. The CSM proposal should be considered holistically, as @dgusakov has pointed out, and cherry-picking its facets and extrapolating disaster scenarios doesn’t seem appropriate (i.e. one should at least consider limitations of TVL, the specified reasons for incentive structures, the fact that I think we all agree pure solo staking is not something economically viable at the scale needed to adequately secure the network and thus needs to have barriers to entry reduced).
In its current form, the CSM would improve the Lido NO diversity, but would have a negative net impact on the decentralized staking ecosystem, as it would incentivize stETH growth at the expense of decentralized alternatives.
This is not true and here’s a simple example: it could be the scenario that 1 year from now if the CSM has been activated the Lido protocol only has 32% of staked ETH. “but that would never happen!” sure it has, Lido was at ~30% in May 2022 and is at 32% now in Nov of 23. Other decentralized LSPs have grown (in absolute terms at least, and some in relative as well), and the number of decentralized LSPs has substantially increased. In this case, the set (both Lido and the network) is not only better decentralized, but there’s no bearing on whether the relative growth that may have gone to Lido has gone to other decentralized alternatives (or CEXs, which seems to be the trend regardless).
You claim that the CSM economic incentives would do so at the the “expense” of decentralized alternatives, but IMO the reasons you’ve outlined aren’t really convincing.
NOs get in CSM get a higher fee than Curated ones
Ok, so what? They will run much fewer validators, so it makes sense.
subsidize APR in order to convert operators from decentralized staking protocols.
I don’t buy this argument, but by this line of argumentation all decentralized LSPs are really just trying to take users from each other, ergo DiVa doing it is OK but it’s not ok if Lido is doing it because the only consideration that matters here about this is whether is someone near 33% or not which is obviously untrue (there’s other things that matter, and might not matter more to you, but they certainly do to me and others as well).
Also, I think here you’re not taking into account the secondary rewards for other LSPs (e.g. have you factored in RPL rewards here?).
As Dima points out, these are not subsidies. The protocol is not incurring losses to provide these increased incentives, it is not changing overall fees on stETH (to make the token itself more attractive), and because this module is permissionless it can be benefitted by anyone; rather it is a reasonable analysis given that these operators would run fewer validators overall (both in mean and median) compared to the curated set, and given that running validators is roughly a fixed cost (or rather starts out ~fixed and then increases with size and complexity but basically only at the professional level) it’s clear that rewards must also be greater for this to be an economically enticing endeavor.
This is further aggravated by the possibility that Lido will provide financial incentives to solo-staker software providers or node clients, which can drive their UX to favor Lido installations over other decentralized alternatives.
Nothing stops other protocols from doing this as well, or creating their own solutions to do so. Monopolistic behavior is characterized when someone is using their position to do things that others cannot, preventing them from doing so on equal footing, or crowding them out. None of these are the case. In this case it is expressly the opposite! If Lido DAO were to support these implementations in open source applications would make it easier for other LSPs to do the same at a fraction of the original cost, since a lot of the code could be re-used across the different LSP implementations. It’s pretty clear if you just apply this “danger” to another similar topic: DendrETH: A trustless oracle for liquid staking protocols , the mutual support in Dendreth by multiple LSPs not only has the benefit of the creation of a solution which could be utilized by multiple protocols (by virtue of being coordinated across a group of participants and thus commonly useful), but also minimizes (or even potentially reduces) the cost of this resource being develop (and possibly used) by competitors to zero.
Looking at the impact the Lido protocol has had on the network, it’s clear that it has meaningfully decentralized it in numerous substantial ways: distribution of stake and preventing single entity CEXes or NOs from amassing outsized stake weight, client diversity (CL and EL), infra diversity, geographic diversity, censorship resistance, sustainable funding of client teams, etc… Where it definitely needs improvement is permissionless participation and a broadened NO base – it’s odd to now say “well, you can’t decentralize there too, that’ll be bad for others!”. Most importantly, large-scale cannibilization of other decentralized LSPs is unlikely IMO. 1) these protocols already have strong communities and often largely sticky NO sets because of their economic models. To wit, they rely on secondary token and/or utility token dynamics which means that the NOs are largely “illiquid” unless they are in a position to not incur serious losses when transitioning, 2) saying Lido has a monopolistic position is something bandied about constantly without the facts to back it up. The protocol does not have a monopoly, only ~32% of stake flows through the middleware (by definition you could have a network with 2 additional participants at this market share) and it is highly liquid stake (i.e. liquidity and withdrawals make it extremely easy for people who disagree with the product to leave it VERY fast). Monopolistic position would be something like 51%+ (in the case that monopolistic acts were utilized) (or 67%+), and using monopolistic powers or mechanism to do things like close the market, crowd out competitors, mandate that node operators who use lido do not use any other LSPs, etc., none of which it does (in fact the protocol is more open and freer in terms of entry and exit than others in these regards). 3) Competitors and self-styled decentralization clergy do not get to declare for everyone else what is decentralized and promotes decentralization or not. I realize we all have different view points, but historically the actions of the Lido DAO and most importantly the impact of the Lido protocol on the network prove that it has been practically and pragmatically one of the largest decentralizing forces on the network. You believe that a Lido capped at 33% will practically decentralize the network further, I believe it will not, and we’ll end up with something far worse.
While this proposal could be a net positive in a world where Lido self-limited, the reality is that Lido continues to increase its dominance with a stated goal to replace centralized staking.
Again, it doesn’t. Lido stake weight is today only slightly above what it was in May of '22.
The size of stETH has already been an extremely contentious topic, with the risk of social slashing looming over Lido. This proposal should not pass without modifictions, as it can damage Ethereum’s decentralized staking and further increase the risk of social slashing for Lido.
This is FUD at best. There’s no real threat of social slashing looming over Lido, and it’s disingenuous to say so. Social slashing is of course Ethereum’s last and ultimate resort against an attacker, but there is zero reason to believe that the Lido protocol (a) could be practically utilized (* by practically I mean "with a reasonable likelikhood and in a way that is actually possible realistically vs just theoretic) to attack the network at current stake share, above 33%, and even above 50%, (b) that that attack wouldn’t be met with extreme prejudice not only by existing stakers, node operators, the Lido community, and the wider Ethereum community itself, and that (c) that there wouldn’t be enough time for an immune response to materialize and be deployed.
Social slashing is a defense mechanism whose main usefulness and utility is as a deterrent against malicious acts in the first place. It is exactly for this reason that it makes such little sense for the DAO to attack the network or for anyone else to try to use the DAO for a vehicle to try to do so. Once the network uses this it will seriously endanger its credibility and ability to use this mechanism in the future against real substantive threats. For a very similar reason things like “MVI” and issuance changes are short-sighted, many see that economic security is “overpaid for” when in reality the “overpayment” creates positive externalities in terms of user confidence and a culture of economic and social security that is much more easily assailed otherwise.
Why is this premature? It’s not realistic or practical to propose a structure when there are so many moving parts that are exogenously determined. The work can progress in parallel and the fee structured can be agreed upon closer to testing. Rocketpool’s fee structure was constantly changing and even changed last minute (and then after it went live). I agree that a fee structure should be put forth clearly before the actual “go live” of the module happens (which would require an on-chain vote regardless), but there’s plenty of time in between then and now.
but decentralizing WITHIN Lido as an entity does not decentralize the beacon chain.
Lido is not one entity, it is a protocol used by multiple entities. It absolutely does and has, decentralized the beacon chain, from all practical effects, and especially from the perspective of staving off centralized stake (which recently and in light of growing institutional appetite has begun to skyrocket again). I guess you can loosely look at the DAO as one entity, but saying “pure DAO == any other entity (e.g. a company incorporated in a jurisdiction)” is plain wrong.
If this economic model is useful in pulling in net new solo stakers to the protocol (which I think most will be), or at least existing stakers running some sub-set of their new validators on Lido as well as existing decentralized LSPs, it is absolutely worth doing. If the worry is that this economic model is better than other LSP models well then they can just come up with better models – besides, there are things that are important besides such APR, and the affordances that RP gives to its NOs (e.g. via the minipool contracts) are a unique feature and selling point on their own, and one that contributors don’t wish to propose to compete on EXACTLY because it has already been done and done well. The point is to make solo staking as economically viable as possible, not to pump bags.
This allows solo stakers to run fault-tolerant validators, improving uptime and effectiveness. DVs enable active-active redundancy, ensuring the validator remains operational even when one or more nodes are offline (as long as the threshold of online operators is met).
