# Roadmap for Eth 1.x into 2.0 execution environment Since this is an area that is being actively researched it seems appropriate to include a DISCLAIMER that what follows is my (`@pipermerriam`) thoughts and opinions and ultimately the path "we" take will depend on the consensus and agreement of the stakeholders in the Eth1.x research group. ## Introduction This document is intended to capture the roadmap laid out in this blog post (currently not publicly published so please do not share link) https://medium.com/@pipermerriam/a-comprehensive-plan-for-ethereum-1-x-7da3af79c418 I propose the following goal for the Eth1.x network. > Changes to the Ethereum mainnet with first class support for stateless clients The term *first class* is intended to signify that the network can healthily support an arbitrary number of stateless clients. This is included to differentiate from the current state of LES light client infrastructure. ## Timeline The rough timeline for this project is 18-24 months. This amount of time was arbitrarily chosen and should be revisited regularly in an attempt to provide the best projections possible for the teams involved. The next 3 months will be dedicated to getting this effort started and establishing consensus within the group on the actual roadmap. Rough plan to be moving full speed within 6 months ## Required Units of Work These are the things that absolutely **must** be done to accomplish the goal. ### Migration to Binary Trie TODO: research to backup motivation for this Binary trie has smaller proof sizes which are *potentially* necessary to reduce the size of witnesses. This is a **significant** task as it requires existing clients to migrate to this new data structure for storage of the state. The specifics of how this is accomplished are still up for discussion but the proposed solution is to have all clients move away from the naive storage of the trie in favor of the design pioneered by turbo-geth which stores the trie leaves sequentially for fast reads and caches the intermediate trie nodes which are generated lazily when needed. This task requires a hard fork. ### Witness generation and propogation We need witnesses to be generated by *someone* and gossipped around the network. TODO: Alexey is working on a proof format. LINK? Beam sync is a likely candidate for allowing experimentation while providing immediate short-term improvements to client UX. ## Optional Units of Work These are the things that very likely make sense to do along the way. ### Beam Sync See: https://medium.com/@jason.carver/intro-to-beam-sync-a0fd168be14a There are two primary motivations for beam sync. First, Beam sync provides a safe testing ground for testing witness generation and propagation. The term "safe" here means that attack vectors found in this area of the networking protocol are likely to effect fully synced full nodes less or not at all. Second, Beam sync should allow clients to more easily join the network. This should lower the barrier to entry for Eth 1.x clients, fostering greater client diversity on the network. ### Ancient chain data pruning TODO: link to @karalabe proposal on how this would be done. Allow clients to prune old chain data. Update status message so clients can broadcast how far back they can serve Blocks and Receipts. Motivation to include this is to allow for clients which only maintain some amount of recent data to still participate in the `ETH` protocol. These clients need a way to broadcast to other nodes what data they should be expected to be capable of serving. This also applies to nodes that are beam syncing. ### Beam Sync protocol upgrades TODO: `@carver` include information about this. Beam sync via `GetNodeData` is sub optimal (by a huge amount, roughly 300x). Beam sync would benefit hugely from some form of improved mechanism for fetching state on demand as well as block witnesses or block witness metadata. Notes: https://github.com/ethereum/trinity/issues/1225 ## Group Structure Here is the proposed structure I have in mind to make this work. I (`@pipermerriam`) will take responsibility for central coordination. This includes scheduling calls, leading discussions when appropriate, and being the primary point of contact for this cross team effort. James Hancock will be assisting us with the Project Management aspects. Alexey will be leading the research component. This will be a collaborative process with the various implementation teams to ensure that the architecture of individual clients is being taken into account and that we come to a path forward that everyone can agree upon. Each client team **should** commit to some level of dedicated resources. I won't try to impose strict constraints on how clients choose to participate but I will strongly advocate for dedicated full time developers who's primary mandate is a focus on this effort. Teams that need assistance in paying for these positions should reach out to me directly `[email protected]`. We will also have periodic in-person summit style meetups 2-3 times annually to come together, share our progress and spread information. Currently the teams that have expressed interested in joining this effort are > (please add your team, I don't want to speak for anyone). - Trinity: primary contact `[email protected]` - EF Research: danny - *your team here*