| Report Date | 2025/01/06 |
| Submitted by | Robert Klotzner |
Polkadot address: 12pRzYaysQz6Tr1e78sRmu9FGB8gu8yTek9x6xwVFFAwXTM8
Current rank: 4
Date of initial induction: 2023/07/16
Date of last report: 2024/06/07
Area(s) of Expertise/Interest: Parachains Consensus
I was together with Donal responsible for the launch of Agile Coretime an Polkadot. The launch was successful, although not trouble-free. Due to the offset of leases only present on Polkadot the migrated leases had been off. A fix was provided in a timely manner, preventing user visible effects of this mistake before they occurred.
I advised on how to make interlacing in Agile Coretime actually resilient and working and I am fixing small issues with partioning in Agile Coretime right now by simultaneously also reducing latency for on-demand coretime by 6s: With #6224 in place an order coming in at block X can already be served at block X+1.
For elastic scaling I contributed with guidance and solutions, e.g. here.
Design wise I have a couple of ideas in the works, super-low-latency messaging will be an RFC in the next period, but for this period I sketched down a way to make parachain runtime upgrades off-chain. This is of lower priority for now, because for reducing barrier to entry we are currently focusing on making smart contracts first-class citizens on Polkadot. On the other hand for teams requiring an actual chain, lower deposits are not their most important concern, but rather latency, reliability and scalability. Nevertheless off-chain runtime play a major role in long-term stability and scalability of the network. It also provides the basis for moving often used smart contracts out of the PoV, further improving scalability.
In the last period, a solution to the probabalistic finality soundness issue, which is summarized here, was found.
I also formalized a general purpose, low-effort sybil resilience mechanism, which should be useful for any p2p network wanting to improve its sybil resilience. It is called Proof of DOT and is useful for many p2p2 protocols, where only a larger number of sybil nodes can do harm to a network, e.g. distributed hash tables. In particular it provides the means to make reputation systems in p2p networks actually work. With sufficiently long lock periods, sustained attacks become costly very quickly. I can also see how this could provide the foundation for implementing full node incentivization mechanisms for supporting efficient light client operations.
In addition to Proof of DOT, a simple mechanism to make the collator protocol more resilient was also devised. This relies on the more generic UMP messages introduced in RFC-103, which is proving over and over again to be a very powerful and enabling technique. It is also the enabling piece of functionality, making very low latency messaging possible.
We identified that one of our short-comings is that we don't actually show off. We think our system can handle a lot of load, because it was designed to, but we never actually tried. This has two important implications:
To tackle both of these, we decided to buy cheap cores on Kusama and put the network to the test. The resulting event is known as the "Spammening" and can be considered successful for a number of reasons:
On a smaller scale I connected with the local Polkadot community and talked about Polkadot and the spammening in particular at the Beefi meetup in Prague.
Experience with the launch of Agile Coretime and the Spammening showcased that we need to get better in quality assurance, making Polkadot the resilient network it is technically setup to be. Further iterations of the Spammening are planned and we are investing in better testing strategies and techiques.
I usually try my best to be quite active and apparently managed to vote in more than 50% of the proposals I could vote on, in total numbers, I voted on 129 proposals according to subsquare.
Threshold