Reevaluating Neuron Control Restrictions

[bjoernek]

@hpeebles

If the disbursement key can only be used to disburse a fully dissolved neuron then it has no impact on people buying neurons who never intend on dissolving them. I’d happily buy an 8 year neuron at a discount regardless of if it has a disbursement key and just leave it locked and earning rewards.

Yes this is a fair point by you and @skilesare.

[bjoernek]

@Seers

will the PoK/disbursement key be inherited by all pieces?

I suppose, there shouldn’t be a way from PoK to non-PoK.

I agree. Once a neuron is in the non-transferable camp (via PoK or for II via a disbursement key), then there should be no way back.

[bjoernek]

@ajismyid

What if someone created a DApp called ‘Neuron Rental Market,’ where neuron owners could transfer their neurons to the dApp controlled-canister, then, renters who access the Neuron Rental Market DApp would have permission to use the voting power associated with those transferred neurons without needing to actually transfer ownership of the neuron between the owner and the renter. Instead, the DApp canister serves as the custody/intermediary. The renter does not care about disbursement or claiming maturity. They just want to use the voting power. Under this system, wouldn’t the disbursement keys lose all significance for protecting ICP governance security?

In this case the neurons under the control of this dApp would be canister controlled and hence the system would recognize them as transferable and apply a discount to the rewards. So the described set-up would be possible but those neurons would get lower rewards.

[dfxjesse]

@bjoernek

Phase 2 (in the way it is currently defined) would not be a big change, from my perspective (but I did not review this with the NNS engineering team in detail yet). Hence I would assume that this could be done in a timely manner after phase 1, subject of course to prioritization compared to other things with respect to NNS/SNS.

Phase 3 would be a relatively big change. We just started the discussion of the high-level ideas and it will take probably take some time to define the detailed design (assuming we reach an agreement to go in this direction). Afterwards it would take certainly several months for implementation.

Will phase 1 be released regardless of whether phases 2 and 3 have been fully defined, or what is the plan? (and what is the timeline for phase 1)

[bjoernek]

@dfxjesse

Will phase 1 be released regardless of whether phases 2 and 3 have been fully defined, or what is the plan? (and what is the timeline for phase 1)

I think as part of the syndication we should agree on the general direction of travel and also confirm this via a motion proposal. I think the motion proposal would not need to describe all details yet, but a sufficient amount of detail so that the majority feels comfortable.

If an agreement is reached, phase 1 (Lift restrictions on canister control over neurons) could be implemented relatively quickly from a technical perspective.

[Seers]

Since everyone agrees on Phase 1 and agrees that there is no immediate threat, maybe we can lift the restriction and gather more data on the pros and cons for a period of time—let’s say one year—before revisiting the discussion with more information and deciding on Phases 2 and 3. After all, that’s the power of the NNS: we don’t need to get things right from the start, and there is a bit of room for experimentation.

[dfxjesse]

Ye, it seems the majority of the discussion is around phases 2 and 3.

We are kind of already in phase 1 with ECDSA and HTTP outcall controlled neurons - So it would make sense to just remove the self-auth principal check (which is an easy change) and proceed the discussion over time as you said with more real world data.

I wonder what’s the counter-argument for proceeding like this?

[bjoernek]

@Seers

Since everyone agrees on Phase 1 and agrees that there is no immediate threat, maybe we can lift the restriction and gather more data on the pros and cons for a period of time—let’s say one year—before revisiting the discussion with more information and deciding on Phases 2 and 3.

I think it would be useful to agree on the overall strategy before we go ahead with Phase 1, but I agree that not all details of later phases have to be defined upfront. This would give some planning security to developers working on projects which involve canister controlled neurons.

With respect to your point on data: I think this is a good avenue and I had similar thoughts. For example if we established some periodic metrics showing the (potentially approximative) breakdown of neurons into different subclasses (e.g. transferable because II-controlled, in the future canister-controlled etc) then this would be very helpful input. However, I think this is not straightforward and I have to think a bit more about this.

[dfisher]

Thank you. But the more fundamental question is why have any non PoK neurons at all? It seems to me best to force all neurons to have PoK as the very existence of non PoK neurons has limited upside for big downsides (added complexity, threat to governance, etc etc).

[LightningLad91]

I think there are very very few neurons that pose a threat today. OpenChat, maybe a couple others? You think we need to force all existing neurons to do a PoK just to cover the few that aren’t already non-transferable?

[dfisher]

I’d like to double click on this idea which I think is great.

1. Making all neurons have a PoK

2. Making all neurons transferable, so I can transfer my hardware controlled neuron to a canister or to II

Regarding 1, I believe that there is simply no upside to having any neurons at all that don’t have PoK. We are planning for the future and we don’t want a large number of non PoK neurons to emerge as those neurons could end up on a marketplace. It is just a bad idea.

Regarding 2, this is simply a brilliant idea. There are a lot of neuron holders out there who created neurons using PEM files and hardware wallets, and frankly, their entire security setup is a joke since it entails putting seed phrases in different places or encrypting PEM files and hiding them. It’s not a professional institutional setup and it’s quite frankly ridiculous. This is true for the vast majority of the voting power on the IC. If people who setup neurons with PEM files could transfer them to Orbit when it finally launches to have an institutional multisig that would strengthen the security of the network as a whole so entire neurons don’t accidentally get lost. And since all neurons are required to have PoK, then we preserve the feature that limits neuron marketplaces. It is the best of both worlds.

@bjoernek thoughts?

[LightningLad91]

I guess I’ll just throw this out there.

I think forcing existing neurons to have a PoK might have unforeseen consequences.

Basically, stakeholders that keep up with current events will immediately register their PoK while others who may not be as active in governance could go months without realizing that they are suffering a penalty.

Please correct me if I’m wrong, but I think this would result in greater rewards for active members and a significant shift in voting power. Maybe that’s ok? Idk.

[dx111]

I disagree with the second and third phases.
What ICP needs is simplicity and user-friendliness. For instance, the experience with Polkadot is remarkably seamless, devoid of the need to grasp intricate terminologies. Anyone can engage and disengage effortlessly without daily voting obligations or apprehensions regarding diminishing rewards. As for security, I harbor skepticism that anyone would expend billions to acquire neurons, only to enact detrimental actions within the community. In this vein, a waiting period akin to Polkadot’s 28 days could be considered.

[Seers]

a waiting period akin to Polkadot’s 28 days could be considered

This could be another option. Perhaps a transfer delay of ~1 month during which you can’t vote, or maybe it could vary according to the dissolve delay (VP).

It seems a good property to allow a neuron to be able to exit/enter the ‘full reward’ state (possibly repetitively), depending on whether it might be transfered (hot neuron) or not (cold neuron). @bjoernek

[dfxjesse]

@LightningLad91

Basically, stakeholders that keep up with current events will immediately register their PoK while others who may not be as active in governance could go months without realizing that they are suffering a penalty.

When I read the proposal I took it as all current neurons are fine and would still get full voting power and only after the cut-off date would the new neurons be subject to the reduction in voting power:

@bjoernek

In order to prevent creating a counterproductive incentive for neuron dissolution among existing neuron holders and to honor the commitment of current neuron holders, it is suggested that the voting power reduction for non-PoK neurons would only apply after a to be defined cut-off date

Does this mean we would assume all current neurons are PoK compliant?

[LightningLad91]

@dfxjesse

When I read the proposal I took it as all current neurons are fine and would still get full voting power and only after the cut-off date would the new neurons be subject to the reduction in voting power:

I agree that’s how it’s presented in the proposal. If I’m understanding @dfisher correctly, he seems to be proposing that we force PoK on all existing neurons too. I just wanted to point out that I think this will have a huge impact on the existing distribution of voting power and rewards.

[alexander]

I don’t quite understand the assertion that ALL NEURONS must be PoK. By “All Neurons,” I interpret it to mean every neuron that currently exists and those that will be created. I find this completely unfeasible. A vast number of neurons have already been established, and many have simply chosen to follow others, stepping back to concentrate on their own matters. Eight years from now, they’ll be taken aback to discover they haven’t received the expected profits, or conversely, could have earned more had they not missed the opportunity to perform PoK at some point. This must categorically not be done!

[skilesare]

@alexander

This must categorically not be done!

This is a governance system and not a structured security. You get paid for governing. If “we” decide to stop neuron following because it is a threat to the network(or some shorter measure of this like periodic confirmation), it is not only our right, but arguably our duty to do so. Why should we eliminate ability to govern the network so people can reap rewards for not participating? I don’t understand the logic behind this at all.

[pixld8ta]

@skilesare

If “we” decide

It is not the first time I read this…

“we” - who is this?

[skilesare]

The NNS majority. (I have to include more characters to post)