What if you had protocols that did not liquidate people? Been following Ruler Protocol recently, they seem to be pulling this off. @vbuterin
1 Like
Governance tokens like UNI seem particularly sensitive to market trends. Without other value propositions, it seems risky to rely on them for Sybil resistance for something like price oracles, where there might be big incentives to act maliciously.
In UNIās case in particular, the market cap has hovered around 300M USD for a long time. Is there not a huge risk for it to return to those levels (or lower) when the market sentiment turns bearish again?
4 Likes
If I am understanding correctly you are assuming a single CDP will be so massive that it will be larger than the market cap of Uniswap? I am envisioning this oracle system to be monitoring massive amounts of individual CDPās that are all being monitored separately not as one large target.
I do not think it is realistic scenario that a single CDP created by a user will be able to generate such a large collateral event to rival a take down of Uniswap. Perhaps limits could exist with how large of a CDP could be created, or governance can vote for certain pairs that can create colleterial safely depending on distribution of the paired token to token holder list? (i.e. if a Link developer owns 10% of tokens of a 50 billion dollar token, then there may be a risk of collateral for a LINK/USD pair as one person could create 5 billion dollars colleterial?)
1 Like
With Chainlink v2 offering a crypto-economic system with staked assets backing its oracle reports, why not leverage both the aggregated data supplied by Chainlink alongside the staking-backed oracle reports to achieve the same ends? You get better data quality than uniswap with the crypto-incentives desired to maintain availability and security.
2 Likes
@vbuterin Can you please address more clearly your issues with why Chainlink does not already meet the tasks youāve outlined?
You write:
- Chainlink is a complex system: What makes this bad or what makes it a negative in being used as an oracle for the stated purposes?
- Incentives are not as clean Please expand upon this - the Chainlink incentives are incredibly straightforward and the cost of an attack using incorrect answers is already way higher than whatever the value of the attack is, and with superlinear staking would guarantee for future purposes that an attack like this will never make sense on the Chainlink network
- There is not an automated mechanism for penalization Aside from this not being factual, why would this be advantageous in a scenario where ā as youāve declared ā time is not of the essence? Youāre also envisioning a system for UNI where a liquidation process āwould be challengeable and could be challenged many timesā - this doesnāt sound automated to me, it doesnāt seem like a lack of automation is a real problem to you.
I am pretty heavily invested in both of these projects and personally would like to see UNI focus on what theyāre doing (building an awesome DEX) vs. taking on an entirely new architectural project for which there is already a massive leader in the space (LINK) which is fully capable of achieving the tasks youāre concerned with.
Not for anything, but after reading some of your replies to the rebuttals the original post comes off very much as if you have some sort of grudge against Chainlink. It would make more sense to be proposing to them how they might make adjustments to be a superior solution for UNIās use case than to ask UNI to just suddenly build a robust decentralized oracle.
5 Likes
lol, the chainlink shills are out in full force in this thread already. i think this is a fantastic idea from vitalik and i trust the team here have the capability to make it happen and function as intended. keenly looking forward to seeing this being developed!
8 Likes
Isnāt this pretty much exactly what Chainlink outlined in their whitepaper 2.0 with their two tier resolution system? For contracts ok with long resolution times, Chainlink will have a second arbitration layer thatās basically the same thing thatās been outlined here with the Augur comparisons.
I guess Iām struggling to see why weāre trying to solve this hyper-focused oracle use case with a governance token for a DEX when we already have an oracle focused project solving it. Competing solutions are great but this kind of feels like a tacked on feature with how specialized it is. Why burden Uniswap with developing this very particular use case thatās orthogonal to their core use case?
Chainlinkās staking system has a quadratic scaling effect, the proposed Augur/UMA model does not. As mentioned in the OP, the Augur/UMA model has an attack cost of 1/2n, significantly less than Chainlinkās n^2 that comes from concentrating the staking rewards from all nodes into one.
So yes, UNIās market cap is massive but LINKās is as well, uses that capital more efficiently by a power of 2, and has better incentives to keep that massive market cap (staking and utility purchasing data) whereas UNIās value comes from governance.
3 Likes
Im curious what is wrong with staying focused on getting ETH moving and having Chainlink solve this issue. Let the leaders of each of these important and thought out systems both doing different and important tasks stay their course.
Whatās your beef here with LINK.
1 Like
I do think it would be great for DAOs to also double up as oracle systems. DAOs make profit and hence have something to lose from being malicious, theyāre also often decentralised which makes coordination for the sake of collusion harder. But I donāt know if Uniswap is capable and willing to work on this, Iād rather see more oracle-focussed project build the infrastructure and then incorporate the UNI token for security.
Securing oracles involves a large number of hard research problems that ethereum researchers are only now waking up to.
- Expectation of profit. Market cap remains high because of expectation of profit. What happens when the expectation of this profit crashes? Equity-like models where tokens earn profit till end-of-life have various valuation models - they all indicate that expectation of future profit can crash much faster than actual profits in the present. Thorchain and UMA have independently suggested changing the fees charged to ease in and out of such crashes, itās a start but not a complete solution yet.
- Competiton may be bad. Such expectation crashes could also means, prima facie, we should try to limit competition between oracle systems because itās the existence of competitors that cause projects to crash. Thereās a reason we donāt have competing courts or judiciaries in the real world. Merging existing oracle systems - chainlink, uniswap, UMA, etc is a way to increase security without competition. But this increases bureaucratic burden just like in the legal system.
- How much value is secured? Itās hard to place a cap on how much value your oracle secures, as doing so breaks some types of synthetics and mechanisms. Unless the oracle is permissioned, it could be securing value without having any on-chain awareness as to which all projects are querying it. In fact, deliberately putting too much value on the oracle to secure is an attack vector because it forces everyone else out of the market because they can no longer trust the oracle. The oracle operator themselves can create such a situation.
- Voter apathy. People donāt operate strictly on risk-adjusted returns, they operate on laziness adjusted returns. Thereās plenty of millionaires who still use Infura + etherscan instead of running their own ethereum node, so itās possible 51% of the market cap canāt be made to show up for a vote even if their wealth depended on it. Bringing everyone online for a vote amounts to a social coordination problem which can, in my opinion, take weeks. Hence the reasonable threshold has to be much lower than 51%, which begs the question, how much?
- Delegation. Above point indicates that we should have mechanisms for delegation. What are the best mechanisms?
- Coordination premium. Using a DAO like Uniswap instead of a single actor has the primary benefit of decentralisation. Corrupting the oracle requires coordination from a number of actors. How do you measure how much additional security this lack of coordination affords us?
- Subjectivity. Dealing with messy real word data involves edge cases. These edge cases too are an attack vector because people are reluctant to vote if they know they will be penalised for offering an unpopular answer to an ambiguous query.
- Data sourcing. Where is data being sourced from, what is itās quality, what is itās uptime, what do you do if that source is not available, and so on. This is probably one area where Chainlink does well.
I do think connecting to the real world is the most signifcant problem that ethereum needs to solve, and I have previously written about the oracle problem (noma.substack / oracle problem, this forum doesnāt allow links). Iād just rather see a more dedicated and collaborative effort from the ethereum community, than rushing it through with a single team who, until now, hasnāt been focussing on this as its primary challenge.
2 Likes
Ruler protocol still requires an oracle.
1 Like
Sir , my village would like to educate you. Chainlink does objective inputs, Kleros would do subjective inputs. If uniswap were to pivot to oracles, they would have to be objective inputs. I personally donāt see any value in uniswap pivoting to the oracle space, it seems to me a waste of resources. I understand wanting to move governance to l2 and give the uni token some utility, but I suspect there are more interesting things you could do with the token to add value rather an oracle.
1 Like
All oracles querying offchain data have edge cases, you can only minimise them, not eliminate.
For instance if Chainlink is querying BTC/USD price from Binance, CME and ftx, how does it respond if there is a large gap between the prices? How does it respond if these exchanges go down? Or if bitcoin gets hard forked - which fork is now canonical for price queries?
2 Likes
Whatever is within the range of the consensusā¦ thatās what blockchain isā¦
forks exist when thereās lack of consensus, thatās the whole point of a fork
2 Likes
You could potentially avoid that significant decrease in overall capital efficiency during normal functioning markets by allowing for an immediate liquidation using a fast oracle like Chainlink, but those fast liquidations must be secured by a bond until the escalation/consensus based oracle is finalized.
Instead of the escalation-based oracle being burdensome to everyone all the time, as it would be if you raise liquidation thresholds, the bonding only only makes the escalation oracle more burdensome to the ecosystem during periods of stress. That burden is aligned with when the oracle is systemically needed to be trustworthy. The increased capital for liquidations and slight risk to the liquidators due to oracle disagreement could be problematic during black swan events without some lender of last resort, though.
I agree that a trusted oracle linking the native asset securing the base layer to the rest of the global economy is a public good of paramount importance. Iād question if something as fundamental and timeless as this should go beyond the responsibility of one protocol that easily wane in its ability to provide the economic security. A coalition of stakeholders (UNI, AAVE, COMP, etc.) could be better stewards and lead to broader adoption and stronger oracle consensus. But, that comes at the risk of being too complex and the project never coming to fruition.
2 Likes
Very smart. Thanks for the suggestion.
Then the nodes that donāt deliver good data lose their collateral. Sergey has talked about defense in depth and how you get as many nodes/sources as possible to make it so redundant that the bad data doesnāt become the consensus. I donāt know all the technical details beyond nodes are incentivized to provide good data, and lose collateral if they are out of the consensus. You can use whatever % deviation you want for the smart contract afaik and how many nodes you want. I would recommend you look at the chainlink documentation for answers beyond that. Iām entirely clueless about how uni would pivot into the space so like I said it seems to me a waste of resources, however Iāll admit that itās beyond my scope on a technical level.
1 Like
Just signal boosting @kenneth here: If youāre interested in writing a spec for an implementation or publishing research to develop this idea, definitely apply to the UNI Grants Program.
5 Likes
Iām not talking about chainlink consensus, Iām talking about bitcoin consensus.
Suppose bitcoin consensus weakens and we have two bitcoin forks and both have significant price. Now if someone queries BTC/USD, chainlink needs to decide which bitcoin fork theyāre actually referring to.
Iām not saying this is unsolvable or even a hard problem, Iām just saying an oracle can never be 100% objective.
2 Likes
data(dot)chain.link/btc-usd