I am the original conceiver and designer of Augurs oracle system. Wanted to add some information not in this thread yet.
– Augur security margin (assuming it is still designed the same) comes from the universe forking where the oracle service is frozen while the two versions of REP compete for market cap. The one with the higher market cap at the end of the competition gets control of the existing markets. This requires an attacker to not just have n/2 of the market cap, but to have enough to inflate a worthless crypto at a price for an extended period of time. I forget some specifics, but assuming all holders of the fake REP sell (as is rational to do) at slightly above full market value of the real REP (as is required for the attacker to win), the cost of attack in actually n. Its possible to extend this into an infinite security margin by having multiple rounds where the attacker must double the amount of money honest holders have each time. This infinite security margin comes at the cost of arbitrary long amount of time to freeze the oracle services.
– Chainlink V2 has serious security margin issues as laid out in the V2 whitepaper. It seemed suspicious when I skimmed the paper originally, but it was not until a writeup made the week in ethereum news making a convincing case that a second tier does not help at all with the security and it is not actually n^2. That said, I am completely confident Chainlink will resolve this design issue. After all, they had to make it in secret with NDAs and a bit too many non crypto university professors. Now that they are able to get wider community input I am confident Chainlink v2 will be a quality product with many capabilities. Makes me wonder though if Chainlink V2s multi hundred page paper with huge ambitions and complexity as well as security margin issue was why Vitalik chose to make this plea to Uniswap instead. He probably just wanted a high market cap erc20 that was not Chainlink to make the request to since a high market cap is a prerequisite to a secure oracle.
– Augurs model can be sped up close to real time if the data at hand is simply a price number because it doesnt require human interpretation. A node can be coded to just automatically contest and stake against incorrect prices with a single block delay, doubling until a fork. Because it can be automated it wont be as slow as Vitalik and others in this thread are assuming. It would even be fast enough (1 block delay almost always) for liquidation, although someone can delay the price update a few blocks in a row by frivolous contesting. This frivolous delay would probably not be able to be more than several blocks though because of the bond doubling each time. Another catch is that all services relying on this oracle would have to be able to gracefully handle the price feed being delayed, or potentially frozen for a long time if a fork happens.
– I agree with the general sentiment that this is orthogonal to Uniswap raison d’être. While it would add value to the UNI token, it is a pretty serious ‘side project’ to do an upgrade that includes UNI tokens splitting in a fork. It may also for security add requirements for UNI token holders to take action in the event of a major contentious event. It may also be a waste of resources if Chainlink delivers a high security margin solution with the ability to contest and lose increasingly staked amounts. I expect that a sped up and automated version of Augur oracle but for price feeds that typically resolves in a single block will eventually exist, and Chainlink will probably take it. Still if Uniswap governance has a huge amount of money to throw around, this has a chance of adding value to the UNI token, and perhaps this is worth doing, but really think they probably have better value propositions.