Hey @lara. I just realized the implications of this detail in the proposal and don’t like it at all. I believe that daily rewards should be computed based on adjusted voting power, not potential voting power. Sleeper neurons should not be negatively affecting the governance rewards that are allocated to everyone. The total daily reward pie should be allocated proportionally to everyone who is actively participating in governance according to whatever definition is established by the NNS. Those rewards should not be diluted by sleeper neurons.
@lara @bjoernek can we please stop claiming that this prior proposal (the one linked in the quoted text “a while ago” and referenced as having advantages) was introduced to stop spam? The spam had already been resolved when that proposal was introduced (it stopped immediately when the proposal weight went from 20x to 1x on the Governance topic). The proposal was introduced due to a hypothesis that removal of the exchange rate proposal topic would create a new incentive for spam, but that hypothesis was never tested. We have a lot of SNS projects today (none of which existed back then) and none of them have a spam problem. The exchange rate proposal topic was a mechanism that guaranteed daily voting rewards to NNS participants and removing it means that there are some days where there are no rewards. Hence, the hypothesis was that spam would be submitted in order to trigger the daily reward. None of the SNS projects have this type of daily proposal, which means none of them have daily rewards, and none of them are plagued with spam. Hence, there is now empirical evidence that this change to ICP tokenomics, which resulted in the largest drop in voting rewards that we have ever experienced, was not even necessary. The only advantage that could have been created by this prior proposal was to create a source of funding that could be used to strategically help advance decentralization without increasing the inflation that is baked into ICP tokenomics design, but that was not in scope for that proposal and hasn’t been pursued since then. Hence, the change has only had a negative impact on active NNS participants by diluting the value they get out of their active governance participation. I really don’t want to see the same design philosophies that were implemented with that prior change also carried over into this periodic confirmation design.
You are right. The original periodic confirmation proposal 55651 did not include voting as a mechanism for satisfying a requirement for active governance participation. The proposal was focused solely on ensuring each neuron owner is making their own Followee choice on a regular basis. I don’t mind the voting provision since I consider it to be an act of intentional governance participation (I have a hard time believing that people who vote manually are not already intentional about their Followees), but I also would support taking it out since it is not aligned with the original proposal.
There is no doubt that staking ICP with WaterNeuron is far less complex and offers higher rewards than staking ICP in the NNS. If someone is looking for a set and forget option, then waterneuron.fi is an excellent solution. The advantage of the WaterNeuron protocol is that it enables people who only care about the investment returns and liquidity of their staked tokens to earn staking rewards while delegating the responsibility of NNS governance participation to the WTN token holders. Hence, if you stake your ICP as nICP, then you earn higher rewards than you would if you stake in the NNS and you can remove your stake from the protocol at any time. However, you don’t have to worry at all about the changes that will inevitably happen in a mutable governance system such as the NNS because there are other people who are actively involved on the governance side who are addressing those issues for you. Hence, every ICP that is staked through WaterNeuron is actively participating in governance as intended by the NNS governance design.
My preference is to start the voting power adjustment at 6 months. I don’t see a problem with the 1 month linear decrease, but it’s easier to think of the deadline for confirmation as twice a year instead of every 5 months. I think most people will think in terms of how to avoid any penalty on their voting power instead of the deadline for complete removal of voting power.
A reset of following for neurons that don’t periodically confirm would be more consistent with the original proposal 55651. I suppose if you did this then you wouldn’t need the adjusted voting power calculations at all and it would also make sense to remove the grace period ramp. You could focus all the development energy on this part of the code instead of creating the new adjusted voting power code that wasn’t part of the original proposal. I like the idea behind the adjusted voting power, but only from the perspective that it removes sleeper neuron voting power from the total voting power calculation, which in my opinion should be applied to both the determination of absolute majority on individual proposals and the determination of daily voting reward distribution. Hence, if we are not going to apply it in both ways, then I would prefer to remove the adjusted voting power calculation from the design scope altogether. As an alternative, you could simply define voting power as neurons with greater than 6 month dissolve delay AND Followees configured for at least one proposal topic. That way lack of periodic confirmation of Followees completely removes voting power from that neuron from the NNS, which addresses the concern about quickly adopting critical proposals. If this were the design, then even the concept of theoretical voting power goes away for sleeper neurons. We should not let sleeper neurons dilute the voting rewards of neurons that are active, so I would prefer that we simply take away voting power from neurons that don’t have a Followee set instead of the design that has been proposed.
Can you point to any neuron that exists in this configuration. This seems highly unlikely.
If these neurons really do exist or you simply want to design for this use case, then perhaps you could create a property of a neuron where the neuron owner can indicate that they intentionally have no Followees set. If they ever set a Followee, then it would automatically undo this setting. If a neuron is intentionally designated as not having any Followees, then that neuron should get full voting power.
Hopefully my position on this has been clarified earlier in this comment. If the adjusted voting power is not going to be used to allocate daily voting rewards, then I don’t want to use the new adjusted voting power calculation at all. I would rather see a Followee reset and voting power be calculated based on the existence of Followee configuration or designation by the neuron owner that lack of Followees is intentional. Sleeper neurons should not be allowed to dilute voting rewards for active governance participants.
I’m more aligned with the hypothesis provided by @bjoernek that 80% of NNS voting power is likely to participate in governance after periodic confirmation is implemented rather than the lower 40% numbers suggested by @Accumulating.icp. When the Governance proposal topic was removed from All Topics and the weight of the governance topic was increased to 20x, it only took about a week for total voting power participation on the Governance topic to increase from about 10% to 45%. This happened very suddenly. Voting rewards were impacted immediately and there was an active education campaign by the community to raise awareness of the change. In that case, voting rewards for participants who didn’t configure their neurons on the Governance topic were still receiving rewards, but at a lower level. My hypothesis has always been that if we want to increase participation further, then there needs to be a mechanism to remove rewards altogether. That was actually documented in the original proposal for periodic confirmation. If people suddenly start to see that they are getting zero rewards, then they are likely to start investigating. Hence I would not be surprised if that translates into a 20% increase over the participation rates we currently see, which is 60%. For the sake of discussion I think it is reasonable to use the numbers that we know today instead of speculating on what they will be. Hence using the 60% participation we see currently on Governance and SNS & Neuron’s Fund seems reasonable to me. At the end of the day, it doesn’t matter. Let me explain.
Below are charts of voting participation on both the Governance and the SNS & Neuron’s Fund proposal topics over the last 1.5 years. All proposals are not included since it is a very manually intensive effort to collect the data, but I believe the data is representative of the trend. I only included proposals where DFINITY voted. The amount of voting power that DFINITY triggers when they vote is evident in the voting trends for each proposal on the dashboard. They own the largest amount of voting power, so the largest jump is always associated with their vote. The blue trend is the total voting power cast on each proposal, the red trend is the voting power that is triggered by DFINITY, and the orange is the voting power that is cast be all other neurons combined. It is evident in these trends that DFINITY triggers just about as much voting power as all other neurons combined. It is enough to reach consensus on any Governance or SNS & Neuron’s Fund proposal even if it requires the full 4 day voting period.
Fortunately, DFINITY doesn’t abuse this voting power and many people in the community find it acceptable. In fact, DFINITY doesn’t own all the voting power that they trigger. The total voting power that is owned by DFINITY has gone down over this 1.5 year period from about 23% to just under 20%. Over the same time, the voting power that is triggered by DFINITY has gone up from about 23% to about 32%. This increase is fully attributed to the fact that the NNS community has gone from about 0-2% voting power that follows DFINITY to about 11-14% voting power that follow DFINITY. This has occured despite the fact that the proposal weight for Governance and SNS & Neuron’s Fund are both 20x and DFINITY has not voted on 9% of all Governance proposals and has not voted on 28% of all SNS & Neuron’s Fund proposals. Hence, the people who follow DFINITY on these topics have lost substantial voting reward over the last 1.5 years. Yet they still choose to follow DFINITY on these topics, which is their right.
Regardless, DFINITY triggers enough voting power to satisfy this 51% attack vector and this situation has existed for the last 1.5 years. I don’t think anyone in their right mind thinks that DFINITY intends to harm the ICP ecosystem. In fact, DFINITY has taken great strides over the last 3 years to help advance decentralization and those efforts continue today. The new Grants for Voting Neurons program is a major step in that direction on the technical topics. There are very few entities in the ecosystem that perform educated reviews and vote independently on technical proposals. CodeGov does this for IC-OS Version Election and System Canister Management and LORIMER does this today for Subnet Management. However, when this new grant program takes effect in the coming months there will be 3 other known neurons that are also performing some of this work. At this time, there is already 9.5 % total voting power in the NNS that is voting independently of DFINITY on technical topics such as IC-OS Version Election and System Canister Management proposals. This is prior to implementation of the new grant program and prior to any concerted effort / incentives to drive decentralization. If the periodic confirmation proposal were implemented today, my hypothesis is that somewhere around 15.8% of voting power would be cast by entities other than DFINITY (based on 60% participation…or a little less if @bjoernek hypothesis is correct and we end up with 80% participation)
| IC-OS Version Election | System Canister Management | |
|---|---|---|
| CURRENT | ||
| Example Proposal | 132149 | 132171 |
| Total Voting Power Cast | 99.6 % | 99.6 % |
| DFINITY Triggered TVP | 90.1 % | 90.1 % |
| DFINITY Owned TVP | 19.7 % | 19.7 % |
| DFINITY Follower TVP | 70.4 % | 70.4 % |
| Other Neuron TVP | 9.5 % | 9.5 % |
| ADJUSTED | ||
| DFINITY Triggered TVP | 84.2 % | 84.2 % (=100-32.8-15.8) |
| DFINITY Owned TVP | 32.8 % | 32.8 % (=19.7/0.6) |
| DFINITY Follower TVP | 51.4 % | 51.4 % |
| Other Neuron TVP | 15.8 % | 15.8 % (=9.5/0.6) |
My prior position on periodic confirmation has been that we didn’t have all the right ingredients in place to make it worth implementing (since spam was solved in other ways). It may still be too soon to implement it, but we are getting closer. We need known neurons that are viable options to follow on technical topics other than DFINITY. There needs to be incentives for them to perform the work of reviewing and voting independently on technical proposals and possibly making technical contributions. There also need to be reasons why neuron owners will be willing to follow someone other than DFINITY. @dfisher outlined this part of the problem very well previously in this comment, but unfortunately there are not any concrete solutions that have been identified yet to address this issue. It does not improve decentralization if changes are made that simply force people to follow DFINITY due to lack of choices. Hence I think we need to be very strategic about when we implement changes that help mitigate the consequences of default following on older neurons. These are the much bigger issues than a 51% attack vector caused by a reduction of total voting power in the NNS. DFINITY already has this attack vector blocked and I see no reason why it would change when periodic confirmation is implemented.
See the table above in the ADJUSTED section for a real world example of how it would change assuming there is 60% of total voting power on technical topics as compared to today. Yes, the percentage of total voting power that DFINITY triggers would go up, but so would the community of active voters that are not following DFINITY. The drop in percentage would all come from default followers of DFINITY. DFINITY would still control the majority of the votes on technical topics, but we would already be getting much closer to decentralization and that is before the efforts that are ongoing to incentivize known neurons or future modifications that could incentivize followers to choose someone to follow other than DFINITY. Decentralization is a transition over time.