Polynya AMA
Q: Outside of IBC, are there any other promising cross-chain messaging standards that you are seeing? If so, why?
Polynya: I’m not an expert on bridges - IMO, only rollup bridges are good enough, so that’s all I know about. However, my semi-informed opinion is that more promising solutions than light client bridges are optimistic bridges and validity proof bridges. (see: Nomad; Ethereum <> Mina bridge by =nil;). These bridges verify full state transitions, so as long as you assume the weaker chain is not compromised, they are as good as rollup bridges. Of course, rollup bridges continue to work even if the weaker chain is compromised, and are as such the only acceptable solution. Once again, obviously all sorts of bridges will continue to exist, but I choose to focus on the best.
Q: What are your thoughts on the potential for trustless interoperability between rollups?
Polynya: There’s tremendous potential for trustless interoperability, and eventually, full atomic composability between rollups. Indeed, this is the best shot we have at a multi-chain ecosystem that doesn’t fragment security and liquidity endlessly. See work by Geometry on this front: Geometry presents: Slush, a proposal for Fractal scaling - HackMD
Q: Outside of Ethereum + layer-2s, what other ecosystem is showing a uniquely differentiated approach that will co-exist and mutually benefit the Ethereum + L2 roadmap? Can be any L1 or L0 approach.
Polynya: Tezos has the best roadmap thus far, with enshrined rollups, enshrined settlement layer and enshrined data availability sampling. They can also have non-enshrined rollups, of course, which live harmoniously alongside and on top of enshrined rollups. This is the only roadmap in the blockchain space I have seen that materially improves on Ethereum’s technically - and I hope it’ll push Ethereum to upgrade the settlement layer to an enshrined settlement rollup sooner.
Q: Do you have a preference for OR scaling solutions like AnyTrust or validity rollup scaling solutions like validiums?
Polynya: AnyTrust is basically like an L1 or sidechain, but does honest minority consensus instead of honest majority, and can failover to a rollup if compromised. This is a significant step forward over traditional L1s. Validiums take things further - as all state transitions are verified. By the way, neither of these are “OR scaling solutions”.
Q: What are your thoughts on zkEVM and when it will realistically be in a position for widespread adoption? And do you think it will win over something like StarkNet that uses Cairo which is optimized for validity proofs?
Polynya: We have various forms of “zkEVM”. Compiling EVM code to a new zkVM, like zkSync 2.0 is doing, is imminent. Another approach is translating EVM code to Cairo, for use on StarkNet. The “native zkEVM” built by Polygon Hermez and Scroll are further away. These will also be much slower to prove initially, and will require GPU provers and take longer to scale up. It’s unclear to me which approach is the winning one. It’s likely the closer you’re to EVM compatibility, the better shot you have for DeFi, high-value NFTs etc, due to the security capital that has been built up. However, for novel applications that didn’t exist before - like on-chain games - it’s possible Cairo is the best solution.
Q: What are your thoughts on Antonio (dYdX) saying that pretty much everyone is wrong about the scalability of rollups and that they currently offer ~30 TPS rather than ~1000. Most L1 protocols made claims about obscene TPS that were not grounded in reality - is the same thing happening with rollups?
Polynya: Eli (StarkWare) later denies Antonio’s claim - see the tweet here.
You should probably not take Antonio’s claim at face value. I don’t know what Antonio means - given dYdX itself has spiked to 50-60 TPS at times (and according to Ethtps.info we have seen 100s of TPS recorded on other rollups) - but my best guess is this is what the provers are currently targeting.
Provers are very easily parallelizable, and also, optimistic rollups don’t have any proving bottlenecks. At the same time, many of today’s rollups are still in their early stages and many are deliberately throttling. But there’s no fundamental limitation for a well-designed rollup to not do 1,000 TPS - this is purely an implementation detail.
Do note that the 10s of thousands of TPS claims are not grounded in reality for both L1s and rollups - after all rollups are simply centralized L1s that are tuned for TPS, but without compromising on security. In the future, we may see multi-sequencer fractal scaling rollups (i.e. many sequencers resolving to one rollup) pull off this feat, but we are a long way away from that.
Q. I've heard a rollup can be secured by multiple L1s. How does prioritization work if the two L1s diverge (perhaps a particularly large reorg, or fork)?
Polynya: This is not possible. It only makes sense for a rollup to be secured by the most secure L1. Adding weaker L1s to the mix will only reduce security. A more plausible alternative is to have the most secure L1 secure a rollup, and then have optimistic/validity proof bridges to other L1s after state transitions are verified on the secure L1.
Q. Are there any projects currently working on rollups secured by rollups you know of yet?
Polynya: A lot of work is happening in StarkNet & Immutable X ecosystems. Though nothing beyond that has been announced yet, I’m sure zkSync, Optimism and others are also working on similar solutions.
Q. With Aztec, I've noticed it's extremely difficult to use their service without going through their web2 site, which tracks IP address and other metrics. Do you see rollups creating centralized data collection risk by having fewer points to submit transactions through?
Polynya: I don’t know about this - surely you can use other frontends?
Q. Realistically when can we expect rollups to scale to the size they can service real demand? We've seen Optimism get crushed in its rollout and Arbritrum had to end Odyssey early. What are the major hurdles preventing performance right now that remain unsolved?
Polynya: To be clear - Optimism wasn’t crushed at all, it was the RPC node that was underprovisioned. They quickly fixed this - Optimism itself functioned just fine throughout. Arbitrum has been deliberately throttling their rollup, but claim to increase these limits significantly once Arbitrum Nitro rolls out (currently in testnet). However, it depends on what sort of throughput you’re aiming for. This is a years-long journey, and the endgame - fractal scaling achieving global scale - is a long way away. I will note that this is the only way to achieve that sort of global scale throughput (millions of TPS) - though it remains to be seen if this blockchain stuff will be in that much demand to necessitate fractal rollup scaling.
Q. When Arbritrum has had sequencer issues it's my understanding that transactions can still be executed bypassing the sequencer but the route for doing so never felt clear to me. Are there projects working on servicing such occasional bypass needs I can follow, and can you provide more colour on that process?
Polynya: The process is users can submit transactions through L1, and anyone can become a sequencer to process pending transactions. I don’t know if this is activated on Arbitrum One yet, but you can look at Fuel V1 and Polygon Hermez for in-production rollups with this feature. I believe zkSync 2.0 will also have a priority queue mode like so. Application-specific rollups can have much simpler emergency exit mechanisms.
Q: In your opinion, given a multichain thesis plays out (extending past Ethereum + Rollup space to include other ecosystems/clusters as defined by Celestia), what is the most promising trusted cross-chain set up today in the form of security/UX and the potential design space with cross contract calls? "Trusted" here being any form of 3rd party which can include a multisig, validator set, trusted oracle/relayer, etc. Some that come to mind include Wormhole, Axelar, LayerZero and Composable finance with its recent cross-chain bridging/contract space. All of which are adopting IBC standards in some way and the cross-contract design space (that extends beyond basic token bridging) feels like it is just getting started.
Polynya: I think trusted bridges are unacceptable, so I don’t know much about them. This is not to say they won’t exist - of course they will - but I focus on the most robust solutions being developed.