Basho, Input Endorsers, and the Future of Scalability

Opens July 30th 2023 with a blackboard talk on Basho, the most misunderstood era, pointing to a Scott Fest presentation on input endorsers by IO's chief architect.

Sets up scalability with a user named Bob sending a transaction that could move an asset, use a smart contract, vote, or register a certificate.

Scalability is really several properties, starting with predictability: whether Bob can predict the cost, latency, and reliability of his transaction, the single biggest driver for enterprise.

The second property, resource growth: a finite-pool system where the pie is divided among more users versus a truly scalable one where the pie grows as users join.

Breaks system resources into three categories, network, data storage, and computation, and says you want all three maximized as users arrive within the lens of predictability.

Introduces trade-offs: as you grow resources you watch your decentralization index, which is why IO built the EDI with the University of Edinburgh as an independent, standardized metric.

The security trade-off, where high-scalability protocols drop Byzantine resistance from one-half to a third or a quarter, and the egalitarian trade-off from a Raspberry Pi to a supercomputer on fiber.

Basho started as a research project to learn which trade-offs are sacred, with the end goal of more network, data, and computation so Bob gets predictability, lower fees, and faster settlement.

Notes that when the agenda started no protocol was a clear winner despite the marketing from Algorand, Avalanche, and Solana, and nothing from 2015 satisfied all the scalability properties.

Lays out the four major scalability items Cardano is working on today: Hydra, Mithril, side chains, and on-chain optimizations.

Hydra is middleware live on mainnet as a smart contract needing no hard fork, moving work off-chain with the same trust guarantees for fast settlement, low fees, and high uptime.

Calls Hydra fast continuous innovation, a big open-source project at hydra.family with around 1,200 Discord members and a public GitHub roadmap moving quickly.

Mithril is the beginning of data availability and proof systems, including roll-ups, with its first version released and an evolution path like Hydra's.

Points to a DCSpark video comparing data-availability solutions across Chia, Polygon, Algorand, Celestia, IPFS, and Ethereum, framing the research agenda for Mithril 2.0.

A Manhattan-Project-style effort of several firms coming together to catch up on roll-ups, with more to announce at the Cardano Summit, spanning network, storage, and proof-generation concerns.

BLS support arrives in Plutus 3.0, and that Mithril will be an essential component of input endorsers alongside more complicated recursive snark and roll-up systems.

Side chains give fast finality, roll-ups, and different transaction models while driving traffic off the main chain, joining Hydra and roll-ups in getting work off-chain.

Optimizations as better on-chain resource use, citing Plutus improvements shipped with the Vasil hard fork that gave DApp developers a 10x reduction in script size.

Argues that with roll-ups, state and payment channels, and side chains for high load, a two-to-five-year window can meet high growth, though Babel fees and tiered pricing are needed for predictability.

Peras is the next Ouroboros after Genesis, a finality gadget for fast settlement both ways between main chain and side chain, with Babel fees and fee markets requiring hard forks on the 2024 agenda.

Mithril is a huge win for wallets, giving full nodes with fast sync instead of a three-day wait and light clients with full-node security, with Galois-led multi-million-dollar teams on data availability.

Side chains let high-volume apps like World Mobile or Midnight run their specific logic while still using ADA and creating network value for Cardano.

Turns to the big second part of Basho, input endorsers, describing the current single-view or shard-view system where everyone shares one version of history, chaotic at the tip before reaching consistency.

Rollbacks, not roll-ups, differ by algorithm: Algorand has fast finality and no rollbacks, while Ouroboros allows probabilistic rollback, so you wait some blocks for settlement based on your risk tolerance.

Moving to a multi-viewed reality with a main chain plus off-chain work reconciled later, using terminology like input blocks and key blocks.

Different stake pool operator groups work on different transaction threads that eventually converge, so more SPOs means more throughput, with Mithril and extended UTXO keeping the familiar trade-off profile.

Compares de-risking the design through proofs of concept to the Manhattan Project's plutonium implosion shell, where John von Neumann's explosive lenses were tested on the shell before adding a plutonium core.

Input endorsers are not needed in 2024 because Hydra, Mithril, and side chains already make Cardano more competitive under the control of individual DApp builders and open-source projects.

The goal is shifting the trade-off from the consensus algorithm's constraint to what the network and data layer can process, enabling a decade of scalability via Ouroboros Leios, with Peras the current high priority.

Well-Typed, Duncan Coutts's firm, did the historical modeling, and that input endorsers must wait for CIP 1694 since they change the trade-offs and monetary policy.

Stresses Hydra, Mithril, and roll-ups do not change what people signed up for with Genesis, whereas input endorsers require rebuilding the incentives layer to pay stake pool operators differently.

Names the goal Omega: all the fast Leios capabilities plus all the good Ouroboros guarantees like Genesis's asynchronous network model, Chronos time-keeping, and self-healing.

A big multidisciplinary project with Stanford's lab, Edinburgh as the nerve center, and Aggelos personally doing the research, with Peras designed over the summer and then Leios.

He wants input-endorser participation possible on a Raspberry Pi or cell phone to leave fewer people behind, adding protocol complexity but a more resilient system.

Pivots to Algorand, distinguishing the corporation Silvio Micali runs, with whom IO has a good collaborative relationship, from the foundation, which poached John Woods and holds an adversarial view of Cardano.

Shows how the Algorand Reddit compresses his adult remarks into a headline about calling Algorand prickly and adversarial, then piles on personal attacks, and notes Algorand has under 10 percent staking participation versus Cardano's 74 percent.

Eight years of research produced roughly 30 of 180 papers strictly on the Basho era, and that Hydra did not exist as code three years ago while Mithril was only a paper two years ago.

Calls IO among the most cited research groups for work like parallel chains, Ouroboros, and Ledger Redux, and says when science cannot give perfection you either wait or make a trade-off, which is why Voltaire and the MBO put the community in the driver's seat.

Solana goes down a lot because it chose a fast-but-brittle trade-off with proof-of-history toxic waste, an ever-growing chain, and few participants who can make blocks, a short-term DeFi bet the market will judge over years.

Cardano has been up over 2,100 days at 100 percent uptime, run by stake pool operators who have never met, something Azure and Google cannot match without central control.

Asks whether you would accept 5,000 to 10,000 TPS that occasionally fails like Solana, saying it is the community's decision, and that Basho first focuses on ubiquitous no-trade-off wins like transactions 10 times cheaper without trusting an intermediary.

Argues good science reduces the scope of controversy, like moving from coal to nuclear power, and urges people off Reddit and Twitter to watch the DCSpark and Ben data-availability videos that talk to them like adults.

Cardano has perhaps the largest open-source research pool in the industry, inspiring ecosystems from Polkadot to Ergo, noting NIPoPoWs just launched on Ergo, from Dionysis Zindros's 2016 PhD idea to the basis of light-client security.

Closes that the next ten years will be more productive than the last, which went from nothing to a three-million-person movement, and that these designs are built to last decades to centuries for nation-states to run their economies.