Pre-proposal: Improve Stake Decentralization

@GRTIQ ’s podcasts provide opportunities for Graph community members to raise awareness to important Graph and Web3 subjects. Last week, Chris Remus raised one such important topic that is very much relevant to the Graph mission: decentralization (EP21 Chris Remus on Keeping Stake Decentralized - YouTube). We often talk about the Graph’s contributions as a protocol in decentralizing Web3. It is also important that The Graph itself is decentralized, which is an important factor to fulfill its mission in the Web3 space. Decentralization hereby extends to many different aspects of the protocol: voting power distribution, open information sharing, staking distribution, protocol control, etc. The Graph has been transparent in its commitment to progressive decentralization, which refers to implementing decentralization in meaningful milestones at point in times when core technical development has advanced and the ecosystem has matured to take the next step in the journey. The most recent launch of The Graph’s permissionless Mainnet is one such example.

As we march forward in our journey to progressively decentralize, it is hence worth taking time to reflect on how well things are playing out in practice. Today, I want to raise attention to one corner in our network, an area which Chris highlighted during his podcast.

The below chart shows the total staking distribution across our Indexer base with each Indexer’s stake shown as a % of the total network. That list includes delegation-focused Indexers, meaning those with a reward cut % of less than 100%.

The key message here is that we currently have five Indexers who represent about 52% of the total network stake (indexer self-stake + delegations) and who represent less than 5% of all Indexer accounts. Without going into deeper discussions on how exactly the chart above should look like in order to represent a decentralized staking distribution, I believe that the above chart is not reflective of that goal.

Besides network security risks resulting from stake centralization, the launch of Curation has now brought further insights to light that demonstrate why stake centralization poses risks to the network. When any one major Indexer allocates heavily on a subgraph, it can change the economic outcome for smaller Indexers in such drastic ways where allocation decisions can be fundamentally altered for smaller Indexers and which may then require allocation changes simply in response to individual moves made by larger Indexers.

Before proceeding with the discussions, I would like to raise an important note. This forum post is not about demonizing large Indexers in any way. To my knowledge, we have not seen malicious and conscious acts made by any larger Indexer in an attempt to capitalize on their network strength. Much of the dynamics are out of direct control for larger Indexers and they merely occur as part of normal protocol operations. When a smaller Indexer makes a mistake, the rest of the network may hardly notice. However, if a larger Indexer sneezes, then smaller Indexers may catch a cold. This does not only impact smaller Indexers negatively, but also creates undesirable situations for larger Indexers who are attempting to establish themselves as reputable members in The Graph community.

Furthermore, I am not advocating to penalize larger Indexers in the form of shifting existing stakes around. Rather, I would like to propose ideas that address the future of our network. We have some 3B GRT currently staked in the protocol and I would like to propose ideas to the question: how can we help the next 3B GRT staked in the network to be more meaningfully decentralized? Chris Remus has mentioned some thoughts in his podcast that had major influence on the three following proposals that I would like to present to the community for discussion and which are aimed to contribute to improve future network stake decentralization:

1. Indexer Decentralization Threshold
The main proposal centers around the idea of establishing an Indexer Decentralization Threshold, suggested at 2%. Any Indexer whose total staking value (self-staked & delegations) is above 2% (currently nine Indexers) would be considered a large Indexer. That would be made visible to Delegators in the Graph Explorer in an effort to stimulate future Delegations to be made with smaller Indexers as follows:

  • The threshold would appear as a line on the Graph Explorer where Delegators choose their Indexer. Above the line are Indexers above the threshold, below the line are Indexers below the threshold
  • The line would be drawn with a message that states: “More delegations above this line can negatively impact network security. Consider delegating to an Indexer below this line”
  • All Indexers above the line are shown in a collapsed view. Solana is hereby a reference example that has introduced this concept in their own form:

The goal for proposal #1 is to raise Delegator awareness to stake centralization at the point in time they make their next delegation and to thus meaningfully influence them to contribute to stake decentralization.

2. New Delegation Rejection for Indexers
Create the ability for Indexers to reject/turn off new delegations. If activated, such a feature would not allow any new delegations to be made to this Indexer. The Indexer itself would be in full control of this feature and it would be entirely their own decision to activate or deactivate it.

Today, large Indexers do not have the ability to directly prevent stake centralization. The goal for proposal #2 is to provide Indexers the ability to proactively contribute to stake decentralization by rejecting new delegations at whatever time they feel appropriate.

3. Decentralization Delegation Tax

Implement a higher delegation tax for large Indexers, suggested at 5x of regular delegation tax (0.5%) which would hence be 2.5%.

The goal of proposal #3 would be to financially encourage Delegators to delegate with non-large Indexer and to thus directly more rigorously influence stake decentralization.

4. Stake Decentralization POAPs (idea added by @Josh-StreamingFast :pray:)

Create different type of POAPs that reward Delegators for considering decentralization in their delegation decisions. Examples are:

  • Delegating to more than 3/5/7/x Indexers
  • First to delegate towards a new Indexer
  • Be among the first 1 million GRT delegated towards an Indexer

We have had network staking decentralization discussions earlier this year led by @Fattox ( Pre-Proposal: Decentralizing Network Stake via Parameter Revision(s) - Governance & GIPs - The Graph) Those discussions too have factored into this proposal and can be considered a progression that takes into account feedback given back then. The goal is to gather broader based support with a compromise that contains less punitive aspects to existing stakes and yet promotes specifically the spirit that I feel we can all rally around: progressive decentralization

The above three proposals are all suggested to address stake centralization, but are not technically linked or dependent on one another, meaning they could all be implemented independently. The key question is whether the community feels that stake centralization is an issue that should be tackled with higher priority in the form of specific proposals. Please vote:

Do you agree that stake centralization is an issue at The Graph that should be addressed and prioritized?

  • Yes
  • No

0 voters


In a second step, I would like to put the above proposals up for a vote. If you agree to the prior question, then here you have the opportunity to express which one (or multiple) of the proposals you agree with the most.

Which one of the proposal(s) would you like to prioritize implementing at The Graph?

    1. Indexer Decentralization Threshold
    1. New Delegation Rejection for Indexers
    1. Decentralization Delegation Tax

0 voters

Later added:

    1. Stake Decentralization POAPs

0 voters


I understand that each of the proposal may require more Forum discussions to nail down some important details. If this post receives broad positive feedback on the above questions, then we would advance with separate Forum post(s) for specific proposal(s) in order to keep the discussions as structured as possible. Therefore, I encourage using this post for more general feedback with regards to the overarching theme of stake decentralization. Also feel free to present your own proposal that would address stake centralization. If feasible, I will incorporate such suggestions into the main post. Thank you!

9 Likes

Great initiative and very happy to see @Fattox ‘s and Chris Remus’ work on this cited.

The threshold piece sounds like low hanging fruit and maybe a quick win. It would be awesome if we could come together as a community and address that one on both the Explorer and the most popular community dashboards at the same time.

I am keen to read more ideas on improving stake decentralization, before I personally tick any more boxes on the topic of priority.

2 Likes

Nice to see this topic coming up again. I didn’t look back on the numbers yet, but I expect the situation has not changed much (for the better), if at all, since my OP which you cited.

As far as I’m aware, the changes to certain variables I suggested back then are still being looked in to, though with what priority I’m unsure. But I’d refer back to those ideas, as I think one or more would have positive impacts for our decentralisation. It’s more a case of where the balance should land for those changes. I’m not sure how the original values were determined, but they’re clearly not working.

Beyond that - market forces would take care of the rest (as well as other stakeholders), imo. I expect Indexers would become more competitive on average, in more ways than one. Higher engagement.

1 Like

The Pareto Principle is a hard one to overcome, that is for sure.

@Oliver, I really appreciate the framing of problem not as how do we redistribute the current delegations, but how do we incentivize the next incoming delegations. :smiley:

My take on the 3 points:

1 - This is a UI issue, and I agree is very low hanging fruit. As having watched a lot of what was happening within Solana before the network really took off, I definitely did notice that when users were helping other users choose validators to stake towards, this point was often brought up. I do believe this a good “nudge” to implement. Although I don’t think it needs much discussion, as it is simply something that a UI can choose to implement or not.

2 - While I like the spirit of the idea, I think a simpler solution that achieves the same outcome lies within the 16x multiple for delegation capacity. Rather than having to code, audit, integrate, and implement a change to allow Indexers to reject delegation (and this would also bring up the questions: is it case by case? does the user spend gas that is lost? does the rejection happen on-chain? etc etc), an Indexer could instead choose to remove some of their own self-stake and instead delegate that to themselves through a separate account. This could “eat up” the remaining room for their delegation capacity and then allow the inherent incentive that is built in by splitting of rewards when over-delegation occurs.

3 - While I like the spirit again, I think there is merit to look at it from the other point of view: change from a stick (higher tax) to a carrot (lower tax). This goes more into a deeper discussion around what brings more benefit to the network: the amount of GRT that is burned through this tax, or greater distribution/decentralization of delegation. I think that Delegators having the option to pay less tax on a dynamic scale would bring greater benefit than a jump from 0.5% to 2.5%. (that is simply a gut feeling, and I more than happy to be change my viewpoint on that). I think there is also less incentive for an Indexer to try and “side step” this jump by splitting their setup into two “medium-large” Indexers, rather than a single “large”. If we attempted to nudge Delegators towards “small” Indexers, then the gas cost of opening/closing allocations, etc, will make this a less attractive path for Indexers (but admittedly not closing it completely).

As for other ideas, I think there’s something to be said about the power of NFTs/POAPs. Delegating to more than 3/5/7/x Indexers? Here’s a POAP! First to delegate towards a new Indexer? Here’s a POAP! Among the first 1 million GRT delegated towards an Indexer? Here’s a POAP!

The cost is nearly 0 to The Graph, but the self-motivation it creates, the ability to show off, etc, these all play on the “let’s do good” side of the brain, rather than the “let’s not be bad” side of the brain.

7 Likes

Thanks Josh! I have added your POAP suggestion to the main post. It doesn’t let me add it as a new voting option of the existing poll (5 minute expiration to modify polls), so I have added it underneath. Great idea!

2 Likes

#3 in the OP didn’t really sit well with me, but I like your thoughts on it, much better to incentivize rather than penalize.

I’d be interested what you think about the suggestions made by myself and others in the old thread that was cited, as the delegation multiplier you mention in #2 was one of the focal points there.

2 Likes

Two observations to add to the discussion.

  1. There are currently 29% of total GRT staked - is that correct? How do you imagine decentralization changing over time as more staked? How can we encourage more staking?

  2. The top indexers all have large self stake and I think fair to assume they are investors who bought their tokens in private sales pre-launch. You see this on every proof of stake network at the start and then exchanges usually start to take over as they slowly launch infrastructure and initial investors sell out of their position.

Usually attempts to limit the size of any single indexer (or validator) are easily defeated by sybil attacks.

I would consider both of these as separate and related issues: how to get more people staking and how to encourage staking across indexers?

I think it is hard to overcome the initial token distribution done via private sales and might be more actionable to encourage the currently 71% of unstaked GRT to stake in a decentralized way.

1 Like

This would be a lot of work, but what if a delegator were able to select multiple indexers in a single delegation transaction (saving gas fees). I don’t immediately see how it would work from the UI side, but arrange the UI in a way where we almost expect people to chose multiple could encourage others to distribute there selection.

Thank you @Oliver that started this topic again.
As I said before, decentralization is good and we, as p2p support decentralization as we can.
I totally agree with #1 and #2 it’s a really good initiative and gives more possibilities for tuning to the Indexers.
I also agree with @Josh-StreamingFast about #3 and totally disagree about #2. Because decreasing delegation capacity by decreasing multiplication numbers doesn’t help anyhow. Because:

  1. Large indexer has resources for creating a lot of indexers, let’s decrease from x16 to x4 and we will get 4 Indexers from huge guys instead of 1 Indexer.
  2. Look at Figment or Framework, both of them have a capacity for 2bln GRT each. Even if you decrease from x16 to x2 they will save the same share of the delegated tokens of the Network. But yes, you can punish p2p, protofire, and some other indexers (including saints “Small indexers” :slight_smile: ).

Speaking about the topic itself… I don’t think so that “centralization” in the current state is evil and any decentralization is absolutely good. It starting looks like a cargo cult.

Below I will describe another view on the situation.

  1. Let’s look at the profitability of Indexers:


    As you can see by the Top 14 (p2p just stared for comparison), some “Small indexers” have estimated APR much higher than “corporates of evil” from the first page by Allocated tokens. Despite on higher Effective reward cut. What a magic :slight_smile: I believe that these numbers already help them to attract “retail holders”.
    So, point #1: Small indexers easily may have higher APR than Large indexers.

  2. Let’s look at “how really” small, part of our small Indexers (actually a lot of Indexers don’t have ENS names and maybe we have much more cases like these):
    Spoiler: I love you guys, all of you are really awesome, I just want to highlight all sides.
    StakeSquid


    201M allocated tokens. With really good realized capacity potential.

Ryabina


85.5M allocated tokens. With really good realized capacity potential.
Point #2: Small Indexers sometimes only seems small
Point #3: Even if you are small but really active and helpful to the community you are growing really fast.

  1. Why some Small Indexers are so profitable, let’s look inside:

    a) Melicious actions, sometimes really bad with possible slashing like in this case or in a case with creating failed subgraphs for themselves. But:
  • self stake is small, so slashing is not a problem
  • doesn’t have any “Name” inside of this or other networks, so, no reputation risks

    b) Indexing “garbage”. Money doesn’t smell, let’s Index everything, even if it doesn’t provide any useful data to Network participants. It allows to make money, and it doesn’t prohibit.
    Point #4: Small indexers could be in many cases only money seekers for the short term. If decentralization has this face, I don’t want this decentralization, give me another, please.
  1. Large indexers in many cases doing helpful work for the Network, for free. Because they are large and they already got some money for getting back to the Network. Some “Small Indexer”, like StakeSquid, WaveFive, Ryabina, Graphops also doing a lot, and it makes them bigger (surprise) :slight_smile:
    Point #5: If you look at things closer, you will see that good actions took their beginning, not from the size of Indexer, but large Indexers take care about their Names and Networks where they work.

  2. We approved GIP2, allowed Indexers to claim their rewards on separate addresses, but postponed the same thing for delegators. It was 4 months ago (!!!). Maybe let’s concentrate on this really important aspect instead of struggle for decentralization without really strong and urgent arguments for it?
    Point #6: Let’s spend time on the most important things first.

  3. All of the large Indexers became large not because of luck. And not because of Sort in Explorer.
    Some of them working hard in other Networks, and their clients from there also joined them here.
    Some were really active on the Network start and attract delegators by their professionalism with answers to delegators and other indexers’ questions.
    Some worked hard with institutional investors, funds and etc.
    Trying to force delegators to leave from these Indexers it’s almost the same that happened in Russia 100 years ago, let’s take away everything from the rich and share it with the poor. Welcome to the soviet union, I’m not sure that all of you are pretty happy to relocate to USSR or Russia, so, why are you trying to create the same environment here?
    Last point #7: If someone got something, perhaps he/she deserved it, let’s learn something from history instead of make the same mistakes every time.

The network already has all instruments for growing to Small indexers, as you can see in the examples above. I agree that we can implement some changes in Explorers, it’s easy and doesn’t make too much disbalance. But please, stop hunting on the witches and trying to find something white\black here, it’s a totally full specter of colors here, don’t punish one network participants, to give more to other network participants.

3 Likes

Neither do I, but I also see where we could end up in terms of stake pooling in few Indexers if we don’t start thinking about how to attract more Indexers to the network and have that stake disperse among more of them. This is an existential threat in my opinion.

Great insight points 1-4.

Point 5 - this is a technical challenge that has already been debated and is generally accepted as desired and has not been implemented because of those technical challenges, whereas we are discussing a pre-proposal - not sure I see any relevance between the two as they are mutually exclusive, even as work items.

On point #6:

I don’t think anyone here is in a rush to ship big indexers off to the gulag, however I agree that there is nuance in the discussion that is always missed in conversation and offhand commentary. It’s not healthy and it’s absolutely not fair commentary on large Indexers that we all see putting in the work. I don’t know how we fix this but it is something I think about daily.

We do have a problem whereby large indexers are assumed evil whereas the reality is most are acting in the interests of both the network and themselves, just like any other indexer worth their salt. In my opinion the problem is the incentives and the parameters we work within on the network and that is what we should be attacking in these discussions; you cannot blame actors for being rational within the parameters and you cannot blame the actors if they are not demonstrating the level of goodwill you want to see in the world, because goodwill is a finite resource and there is never enough to go around.

On the other hand, I also think this “we pulled ourselves up by our bootstraps” mentality is not useful in this conversation, and we’re all guilty of it to some degree. Success is a function of many things including hard work, privilege, luck/fate etc. so any argument that an assumed economic position in a network is a pure function of hard work/capital is not useful. It has no bearing on where I think we need to be - widely distributed stake across a diverse range of indexers.

We all need to be demonstrating a level of humility, no matter our economic weight within the protocol. My concern when it comes to stake centralization has nothing to do with “taking from the rich” - it’s all about how we can bring more participants into the network and have the stake naturally disperse. Punitive actions are completely off the board as far as I am concerned, but when the exchanges come, if we haven’t figured this out we are destined to fall into the same traps most PoS network fall into.

I feel like #6 was the nuclear option in terms of comparisons @KonstantinRM but I still appreciate that you made it. I don’t think retroactive changes that are punitive based on stake will be healthy for the network.

2 Likes

This topic is really interesting and as we’re seeing everyone has a valid point.
But i don’t think the approach is how to limit the capacities of large indexers, in my opinion it’s more about how to enable small/medium indexers.
As said above, small indexers that contributed to the protocol aren’t that small anymore. Small indexers have higher APR than large indexers, and if their “name” is known, delegators will delegate to them, no wonder.
When we delegated the 5M to small/medium indexers, we were saying in another way, that these indexers are trusted by Figment. So when someone wants to delegate to Figment then sees that these smaller indexers have higher APR than us and trusted by us, they delegate to them.
In this way, we’re kind of re-directing some of our delegations towards them.
I think that is kind of the approach, to every while, publish a list of small/indexers that contributed to the graph and shed the light on them.
It’s something at Figment we want to do.
But if small/medium indexers doesn’t want to be active, and doesn’t want to contribute and also forget about their allocations, there’s nothing we could do for them. We’ll be sending delegators towards risky indexers, who might change their cut or just not be a good indexer.
In short: Let’s find a way to enable small/medium indexers that contribute to the protocol and are active

4 Likes

Really good point! Love it.
Will think about delegating to small\medium Indexers, who help the community, part of our profit and share good words about them. Thanks for the idea!

3 Likes

Thank you, Jim. Your points are really wise, as usual :slight_smile: My #6 it’s obviously some kind of willful exaggeration, but I’m sure that we need to enable Small\Medium Indexers without retrospective actions and other potential direct unbalanced activities against large Indexers. Because growing it’s natural event for everybody who works well.
I believe we will get more new Indexers here and today’s Small\Medium Indexers also will grow by new delegators, more competitive landscape will be good, but without synthetic barriers for the growth.

3 Likes

I think something helpful here would be to reframe the discussion entirely, while still holding the same issue up to the light.

As I read it currently: “There are a few large indexers who have an enormous amount of self-stake and delegation on those addresses”
As I think it should be read: “As a network, we want more Indexers to join and serve data. How do we encourage Delegators to delegate to those Indexers as well, and not just on the same large incumbents.”

Both of these points encourage the dispersion of delegation (and thus indexing rewards), and both discourage overly large concentration of GRT flows. But I’d argue that the second version ensures that we tackle this problem as a community looking to grow, rather than as a fragmented “us versus them”.

I am of the mindset that we should always be looking to solve a perceived issue through incentives, rather than through punishment (when possible). Let’s try to nudge and encourage “good” behaviour, not punish “bad” behaviour. If the system allows for “bad” behaviour, is it bad? Or is it just a portion of the community’s view that it is bad? We must never lose sight of our own personal and cultural biases, and that 2 people from different parts of the world will see the same issue in a completely different way.

From the work that StreamingFast has done on other networks, we’ve seen things “devolve” into “extractive” behaviour, yet the network still runs perfectly well and continues on. What seemed extractive and unthinkable to some, was the clear rational path forward to others. Who decides who was right?

As was mentioned briefly, we’re yet to see exchanges get involved in a meaningful way, and that can change the dynamic drastically.

So rather than figuring how do we limit, I think we should focus on how figuring how we encourage new participants to join (Indexers and Delegators alike).

1 Like

I find a lot of what you wrote focuses on extremes or outlier examples. This conversation is not about zero sum going against large stakeholders.

  1. You mention that larger Indexers could just “spin up more nodes” if there was a smaller delegation ratio or cap was in place.

This is true, but you’d also have fairly linear scaling costs for infra and gas especially. Isn’t it better to have a diminishing incentive, which may give rise to large stakeholders considering delegation, over spinning up more Indexers?

It’s somewhat the inverse of how small stakes are cornered out of the market and can’t even consider to allocate to many subgraphs, if they want to make a profit. Which end of that spectrum do you think we should be worried about more?

“Despite on higher Effective reward cut. What a magic :slight_smile: I believe that these numbers already help them to attract “retail holders”.”

This ignores the ACTUAL reality of the delegation weights of those you’re believing to attract these delegations. And yes, it would make sense to delegate to those with higher APY, but clearly that’s not what’s happening.

  1. You make a case that some Indexers who run multiple nodes are doing ok in aggregate terms, but they are not the common example.

  2. You then nullify point #2 by saying “these people are only doing well because they are being malicious”. Which is not only incorrect, but it’s contradictory to the point you were trying to make.

(Apologies for the edits, this forum isn’t that friendly to use on mobile)

3 Likes
  1. No, it isn’t. Maybe I didn’t understand you right way. But, if you already have big self stake you can’t withdraw part of it and delegate to someone.
    If you speak about delegator’s tokens they are not Indexer’s, so Indexer can’t decide to delegate someone’s tokens anywhere. Indexer can recommend his delegators to delegate to other Indexers. It will be almost the same if you will recommend your employer to send your salary partially to me. I’m agree BTW.
    Delegators may decide by themselves, but in this case they should have some strong reasons for that. Not only sake of decentralization itself.
  2. I didn’t make my own research on it, but logically it should help them, share your research please if you did some.
  3. I’m sure that all Indexers shouldn’t get the same profit. If someone who just bought 100k GRT and created a new Indexer by Payne’s setup and does nothing after that will get the same profit as Payne, it’s just totally unfair.
    So, yes, I personally think that it’s ok, that it’s not a common example.
  4. No, it isn’t. I provided several examples, one of them was malicious, other works with not prohibited subgraphs, and I also can provide proofs without garbage subgraphs with higher APR. Just because some small indexers working smarter and harder than others.

Again, I’m not against Small indexers, I just told that some Small indexers get a good profit, some use some cheats, so it’s not about Robin hoods. Let’s keep the system balanced in our attempts to do it a little bit more friendly for one part of the community :slight_smile:

  1. Yes that’s not what I meant. I meant at some point, it makes more sense for a large stakeholders to delegate some of their weight, than to create indexer #2, #3 #10… IF the economics by such changes, means it makes sense. Could they self delegate? Sure, some may even prefer it (lowered risk), but this still ends with other equity having a bigger distribution. You can absolutely do it either way, if we consider balancing these variables based on the future of the network, and thus the absence of the vesting contracts many Indexers and presale buyers find themselves with today.
    I’m yet to hear how the current delegation ratio (or absence of a cap) makes sense at all. It will never begin to until the context changes, because the supply won’t adjust that fast, and by ‘context’, I mean the size of certain individual stakes. Should we wait for that day, or balance based on the current and more immediate future?

  2. The relevant numbers are in the pic you posted yourself, when you spoke regarding APR.

  3. I agree, but are we concerned only with the few MVPs, or the overall state of the network? Should we aim for 20 Indexers, maintained by 5-10 operators? You provided someone like Payne as evidence of “the system is working”, but he’s a (well deserved) exception to the rule and people like this would still thrive far above others even if delegation was distributed more widely. Maybe even moreso in his case, since by your logic, more nodes would be created and such Operators -for-hire would be paid per operation, right? :slightly_smiling_face:

  4. Firstly, you should read the latest arbitration charter before being so quick to pass your own judgements.
    Secondly, you imply that they are working smarter, but “smart” in your examples seem to be bad actions that don’t benefit the network. So we could take this two ways.

  • Either they working smarter and/or harder than you, but still making a small piece of pie, because a higher % from their much smaller weight is so much lower than yours.
  • Or… if they only served “good data”, as defined by you, and not by Curators (???), they would earn even less still.
    I’m not sure what point you’re trying to make here, but I can’t see an upside in favour of your argument no matter how I take it.
2 Likes

I read arbitration charter, but I don’t understand what do you referred to. I think it will be helpful for everyone if you describe it in more details, thanks.

Smarter it’s not always “bad actions”, for example, monitoring subgraphs with small amount of signaled tokens, but real subgraphs and be only one who will allocate to it is it “bad actions”? I’m sure that it’s a good actions and it’s smarter than just allocate as majority by proportion.

I don’t want to argue with you anyhow, especially because we speaking about the same just with a little bit different colors. Thanks for sharing your thoughts :slight_smile:

Stake Decentralisation is a critically important topic, and I’m glad it’s been brought back to the fore.

Having read through this thread, I’m struck with the difficulty of decentralised governance. Inevitably, you have players coming to the table with noise, straw-man arguments, and attempts to derail the discussion away from the core issue because inaction benefits them. This can be an effective tactic, which is why I’m calling it out. I hope that we can, as a community, see this for what it is, and bring the focus back to where it should be.

I am fully supportive of measures to increase the decentralisation of stake within the network. Centralisation of stake is an existential threat to all decentralised networks, and particularly those that leverage token voting for governance. Stake = control. Furthermore, the auto-compounding nature of stake can be unintuitive, and accelerates centralisation with exponential pace over time.

Many great options have been proposed in this thread already. I’d like to highlight that initiatives to increase decentralisation can operate within two layers: social and protocol. Initiatives can:

  1. Create social cues and pressures to encourage stake decentralisation
  2. Create protocol-level incentives (and disincentives) to encourage stake decentralisation

For example:

  • Indexer Decentralization Threshold (UI updates to encourage users to delegate to smaller indexers)
  • Stake Decentralization POAPs (awarding POAPs to delegators that delegate to smaller indexers)
  • Large Indexers publicly declaring a list of smaller Indexers that they are delegating to (lending credibility to smaller Indexers to encourage delegations)

Are all initiatives that target the social layer. And:

  • Decentralization Delegation Tax (varying the delegation tax based on the Indexer’s size)
  • New Delegation Rejection for Indexers (allowing indexers to reject new delegations)
  • Changes to Delegation Cap (reducing the delegation cap multiplier for Indexers)

All create incentives within the protocol that may change the behaviour of various actors to achieve better stake decentralisation.

Personally I’m of the opinion that we need to tackle both layers. The social layer, and the culture prevalent in the community, is really, really important. Any steps we can take to highlight the importance of decentralisation, and instil it within the community as a value that truly matters, should be taken. The long-term value of this protocol and network depends on it.

Any notion that “centralisation is not bad” is absolute nonsense. Centralisation is unequivocally bad in the context of a decentralised protocol and anyone who disagrees is probably unconcerned with the mission of The Graph (and the wider crypto ecosystem) and probably more concerned about maximising their own short-term financial return.

However, depending solely on the social layer to achieve this is a mistake. Culture and community matters very much, but economic actors will always be profit-seeking. Some are able to recognise that the value of the network (and thus everyone’s stake within it) will be much higher if it is credibly decentralised than if it is not, but others cannot see this far, or are more concerned with short-term profits. For this reason, I believe we must also take action within the protocol, to create strong incentives/disincentives to encourage stake decentralisation. Talk is cheap. Incentives drive behaviour.

So, onto solutions:

I’m still noodling on much of this, but some initial thoughts:

  • Fully supportive of implementing Indexer Decentralization Threshold (UI updates to encourage users to delegate to smaller indexers), as well as any other social cues we can implement to highlight and encourage the value of decentralisation in the community.
  • Mechanisms that would force larger indexers to split their stake into smaller Indexers, or ideally delegate that stake elsewhere (like self-stake limits, or delegation cap changes) will be effective. Yes, some of the challenge can be automated (requiring investment), but this still forces a scaling of operational complexity, gas costs and labour costs. These disincentives will have some level of efficacy and while some larger indexers may follow the sybil attack route, it will not be without cost + diminishing returns. @Fattox nails it here.
  • I think a variable delegation cap multiplier may be interesting (e.g. <1M self-stake = 20x. 1B self-stake = 0.5x). It doesn’t directly subsidise smaller indexers, but creates capacity for them to attract delegation and use a positive effective cut to increase their proportionate self-stake in the network faster than the massive indexers, and only if they are able to attract that level of delegation (i.e. demonstrate they are quality indexers to delegators). This mechanic enables good, small indexers to overcome the inertia of the massive compounding stakes of large indexers and actually grow their proportionate stake in the network over time.
6 Likes

I’m going to think over your post a little more and let others reply, lots of good points there. But I just wanted to highlight this part and say that this is a very interesting idea.

“I heard you like bonding curves, bro. So i put a bonding curve on yo’ ratio!” :sweat_smile: :ok_hand:t2:

4 Likes