In July I spent three weeks in Zurich at the Deep Dive into Blockchain summer school. I had an amazing time. I also came home with a narrow opinion about where this technology has value in my eyes. It is not the public chains, it is the permissioned ones.
Three weeks in Zurich
The school is run by the UZH Blockchain Center. Three weeks in July, built on three pillars: the technology, the economics, and the law. Plenty of lecturers rotated through the weeks, split between academics and people who run these systems for a living, and there were enough hours of class that you really feel it by Friday. Hearing a cryptographer, a regulator, and a CTO describe the same system in one week does something to your brain, but even more so, your intuition.
What I actually took away
By the second week I had picked up a pattern. The projects I most appreciated that were in production, with real money behind them, kept sharing one property. The participants in the chain were known to each other, at least in some fashion.
One such example is a permissioned blockchain. A governed set of members, each holding a cryptographic identity, writing to a shared ledger that none of them owns. Another might be a public chain, but with a curated set of validators, so that the operators of the chain know who is validating and can hold them accountable.
Public chains spend enormous effort on one question: how do you agree on state when anyone can join and anyone might be lying? Proof of work answers it with electricity, proof of stake with capital at risk, and both vary in fees and latency. Knowing who the participants are makes the question a lot cheaper.
Hyperledger Fabric
Hyperledger Fabric is a permissioned blockchain I have come to appreciate. Open source, under the Linux Foundation, and running inside large organisations for years. A few of its design choices stand out to me.
- Identity is the foundation. Every participant holds an X.509 certificate from a Certificate Authority, and a Membership Service Provider maps it to an organisation and a role.
- Endorsement policies decide who signs. Require two named entities, or any majority.
- Channels partition the ledger. Members on a channel, which is a like a private subnet, share a ledger and everyone outside sees nothing. When two of them need a price kept from the rest, a private data collection sends the payload straight between them and writes only a hash, enough to prove what was agreed.
- Chaincode is normal code. You write it in Go, Java, or JavaScript. There is no new fickle language to learn and no gas metering to reason about.
- Execute, order, validate. Fabric runs a transaction and collects endorsements first, orders it second, and validates it against current state before committing. Ordering runs on Raft, or SmartBFT since v3.0 for Byzantine fault tolerance (if you're into that).
There is no token, no mining, and no gas. The apparatus a public chain needs in order to bribe strangers into behaving themselves is absent, because Fabric knows everyone.
The XRP Ledger
The other system I came away impressed by is the XRP Ledger. It is public and open to anyone, and it interested me because it leans on a known set of participants, the validators. Validators are selected through a Unique Node List, so each operator picks which validators it trusts, and in practice those are named organisations: universities, exchanges, businesses. The ledger closes every three to five seconds and charges a fraction of a cent, and that fee is destroyed rather than paid to anyone.
XRPL is also narrow by design, which I see as a feature. Escrows, payment channels, issued currencies, and an order book exchange are native transaction types. You configure them rather than deploying a contract and auditing it. Fewer moving parts means fewer places to lose money. You can also do no-op transactions, which is a neat trick for timestamping and proving existence of data.
I used the XRPL EVM sidechain for the escrow contracts in Nkadime, a lending platform I worked on in my first year of my masters.
In February this year XRPL activated Permissioned Domains. Accounts can now be gated behind on-chain credentials, so a regulated institution can operate inside a section of the public ledger where every counterparty has passed KYC. An open chain grew a permissioning layer because the institutions that wanted it needed one. To me that says more than any marketing blurb about decentralisation.
Why do I have a preference for these two?
My position after Zurich is simple, blockchain does its best work when a group of organisations need a shared record, they trust each other too little to let one of them host it, and they all know who the others are. Fabric serves that case directly. XRPL exists as a version where the record is money moving between them.
The rest of the space is interesting to read about, and I'm so grateful for the insights gained in Zurich.