github.com/gnolang/gno@v0.0.0-20240520182011-228e9d0192ce/docs/concepts/tendermint2.md (about) 1 --- 2 id: tendermint2 3 --- 4 5 # Tendermint2 6 7 **Disclaimer: Tendermint2 is currently part of the Gno monorepo for streamlined development.** 8 9 **Once Gno.land is on the mainnet, Tendermint2 will operate independently, including for governance, 10 on https://github.com/tendermint/tendermint2.** 11 12 ## Problems 13 14 * Open source is open for subversion. 15 * Incentives and mission are misaligned. 16 * Need directory & forum for Tendermint/SDK forks. 17 18 ## Partial Solution: adopt principles 19 20 * Simplicity of design. 21 * The code is the spec. 22 * Minimal code - keep total footprint small. 23 * Minimal dependencies - all dependencies must get audited, and become part of 24 the repo. 25 * Modular dependencies - wherever reasonable, make components modular. 26 * Completeness - software projects that don't become finished are projects 27 that are forever vulnerable. One of the primary goals of the Gno language 28 and related works is to become finished within a reasonable timeframe. 29 30 ## What is already proposed for Tendermint2: 31 32 * Complete Amino. -> multiplier of productivity for SDK development, to not 33 have to think about protobuf at all. Use "genproto" to even auto-generate 34 proto3 for encoding/decoding optimization through protoc. 35 - MISSION: be the basis for improving the encoding standard from proto3, because 36 proto3 length-prefixing is slow, and we need "proto4" or "amino2". 37 - LOOK at the auto-generated proto files! 38 https://github.com/gnolang/gno/blob/master/pkgs/bft/consensus/types/cstypes.proto 39 for example. 40 - There was work to remove this from the CosmosSDK because 41 Amino wasn't ready, but now that it is, it makes sense to incorporate it into 42 Tendermint2. 43 44 45 * Remove EvidenceReactor, Evidence, Violation: 46 47 We need to make it easy to create alt mempool reactors. 48 49 We "kill two birds with one stone" by implementing evidence as a first-class mempool lane. 50 51 The authors of "ABCI++" have a different set of problems to solve, so we should do both! Tendermint++ 52 and Tendermint2. 53 54 55 * Fix address size to 20 bytes -> 160 is sufficient, and fixing it brings optimizations. 56 57 58 * General versionset system for handshake negotiation. -> So Tendermint2 can be 59 used as basis for other P2P applications. 60 61 62 * EventBus -> EventSwitch. -> For indexing, use an external system. 63 64 To ensure Tendermint2 remains minimal and easily integrated with plugin modules, there is no internal implementation. 65 66 The use of an EventSwitch makes the process simpler and synchronous, which maintains the determinism of Tendermint 67 tests. 68 69 Keeping the Tendermint protocol synchronous is sufficient for optimal performance. 70 71 However, if there is a need for asynchronous processing due to an exceptionally large number of validators, it should 72 be a separate fork with a unique name under the same taxonomy as Tendermint. 73 74 75 * Fix nondeterminism in consensus tests -> in relation to the above. 76 77 * Add "MaxDataBytes" for total tx data size limitation. 78 79 To avoid unexpected behavior caused by changes in validator size, it's best to allocate room for each module 80 separately instead of limiting the total block size as we did before. 81 82 This way, we can ensure that there's enough space for all modules. 83 84 * Remove external dependencies like prometheus 85 To ensure accuracy, all metrics and events should be integrated through interfaces. This may require extracting client 86 logic from Prometheus, but it will be incorporated into Tendermint2 and undergo the same auditing process as 87 everything else. 88 89 * General consensus/WAL -> a WAL is useful enough to warrant being a re-usable 90 module. 91 92 * Remove GRPC -> GRPC support should be plugged in (say in a GRPC fork of 93 Tendermint2), so alternative RPC protocols can likewise be. Tendermint2 aims 94 to be independent of the Protobuf stack so that it can retain freedom for 95 improving its codec. 96 97 * Remove dependency on viper/cobra -> I have tried to strip out what we don't 98 use of viper/cobra for minimalism, but could not; and viper/cobra is one 99 prime target for malware to be introduced. Rather than audit viper/cobra, 100 Tendermint2 implements a cli convention for Go-structure-based flags and cli; 101 so if you still want to use viper/cobra you can do so by translating flags to 102 an options struct. 103 104 * Question: Which projects use ABCI sockets besides CosmosSDK derived chains? 105 106 ## Roadmap 107 108 First, we create a multi-organizational team for Tendermint2 & 109 TendermintCore/++ development. We will maintain a fork of the Tendermint++ repo 110 and suggest changes upstream based on our work on Tendermint2, while also 111 porting necessary fixes from Tendermint++ over to Tendermint2. 112 113 We will also reach out to ecosystem partners and survey and create a 114 directory/taxonomy for Tendermint and CosmosSDK derivatives and manage a forum 115 for interfork collaboration. 116 117 Ideally, Tendermint2 and TendermintCore/++ merge into one. 118 119 ## Challenge 120 121 Either make a PR to Gaia/CosmosSDK/TendermintCore to be like Tendermint2, or 122 make a PR to Tendermint2 to import a feature or fix of TendermintCore.