Hacker News Reader

Comment by viraptor | Hacker News Reader

It's an interesting article, finally listing some specific use cases. But I'm curious - why now? What makes blockchain better than using other possibilities between each pair of banks? I don't expect bank A to care about transactions between banks B and C, so the distributed blockchain doesn't sound very useful. On the other hand, I expect banks to care about guaranteed atomic transactions rather than consensus.

(Transfers between many banks in the UK were instantaneous for a decade or so, so apparently it can be done without a blockchain ;-) )

View on HN

All the transaction stuff is useless. Blockchain is actually really bad for that: it utilizes a centralized ledger.

Money today is completely decentralized. There is no central registry of where every single piece of money is, which has to be updated every time anyone makes a transaction. It flows and is converted to goods and vice versa in a completely decentralized and parallel way.

Remember: the ledger is completely centralized (there is a single, unique, monolithic ledger). Only processing of the ledger is distributed.

It is probably the worst possible way to design a currency.

This is untrue, where did you learn this?

What is untrue? That bitcoin has a single central ledger?

Duh, it does. If the ledger is processed in a distributed way across many peers doesn't change the fact that there is a single central ledger. If a transaction isn't in the ledger, it doesn't exist.

If I give you $10 and we don't tell anyone about it, do you still have the money? Yes.

If I give you BTC10 and we don't tell anyone about it, do you still have the BTC? No.

I believe you're thinking of the ledger as a distributed ledger and applying 'centralized' as a description, which it is not. Bitcoin is a decentralized ledger.

No, it is a centralized ledger (there is only one unique, central ledger that can't be partitioned) which is processed in a peer to peer way.

Compare this to actual currencies, that have millions (or even billions) of independent, parallel, distributed, ledgers.

Or compare this to DNS. There isn't a single, unique, central DNS record, it is a system of distributed records across he entire internet.

You would benefit from broadening your understanding of centralized and decentralized as it applies to the blockchain and ledgers.

If I give you a paper wallet or something like openDime? You now have BTC and nobody knows about it.

If you give someone a paper wallet, then both of you have access to it and both of you know about it. (You can't prove you don't have a copy) If you do it on opendime, then everyone with access to opendime know about it. (Potentially - depends how much you trust opendime)

Regarding OpenDime, I found these two relevant exerpts browsing the site, which were easy to find.

>How do I know the manufacturer doesn't know the private key?

>When you first plug in a new Opendime, it has no private key. It shows up as a writable drive and the user must copy files onto the drive. They are immediately forgotten, but the the file contents are hashed (SHA256). It's the hashed output of that which forms the private key (along with the unit serial number and entropy we derive from the environment). That happens as soon as 256k worth of bytes have been written to the drive. The private key is created and the drive "ejects itself" and comes back read-only and in "normal" mode... only the public key is shown until unsealed.

>Is the private key unique and secret?

>Yes. Opendime is delivered without any private key. You must give it entropy (random numbers) the first time you use it. Once it's gotten enough numbers, it will hash them all together and use that to pick a random number to use as the private key. At that point, the payment address is generated and set in stone.

>This whole process is very easy: just copy some files into the USB drive. When it's got enough bits (256k bytes) it will eject itself and come back with its final payment address.

Each bank/financial institution will probably have its own blockchain for its own purposes. But if they need to make a transaction to another bank's blockchain, they can use the Interledger protocol [1] to do so.

[1] https://interledger.org/

Here's the point though: if you want to do this, why not have a standard relational database at each bank and a standard for authorised endpoints in each paired bank? This tech was available for decades. Why interledger and blockchains to complicate it?

You wouldn't. You use a blockchain when you want one common ledger across multiple companies, and you don't want to pay/trust a single company to manage it.

Assuming you know who everyone is, you can use traditional consensus mechanisms like Byzantine Paxos and get much higher transaction rates than public blockchains can do right now.

You use a blockchain when you want one common ledger across multiple companies, and you don't want to pay/trust a single company to manage it.

Not really, you can have a blockchain managed by a single company. In the most basic definition, a blockchain is simple an authenticated linked sequence of records. It can be public or not, it can be processed centrally, distributed across a single organization, or distributed in a peer-to-peer fashion.

I guess so but usually people mean something that includes a distributed consensus algorithm.

With bitcoin, because the participants aren't identified and there isn't any form of traditional trust established between them, there are certain actors in the system that are paid a lot to manage the transactions/blockchain.

In your example of multiple companies coordinating on a single blockchain, who would be paid to manage it?

It's way cheaper than on Bitcoin because there's no PoW. Each participant would pay to run their own servers that connect to it.

Why would you want a consensus? In what financial situation would you expect the bank to say: this entity says the account has X, but those two entities say it has Y, so I'll accept Y?

That's not how blockchain consensus works. Once block number X is finalized, everyone sees the same data in block number X.

If a block isn't finalized, and you build a transaction based on particular conditions in that block, and the finalized block has different conditions, then your transaction just fails.

> everyone sees the same data in block number X

The same data in a specific block, but not everyone necessarily agrees that the same block is the latest one. That's why you don't have 100% confidence in a single block confirmation.

Exactly, what I mean by "finalized" is that everyone agrees and there's a very low probability that they'll change their minds.

And until it's not confirmed enough times, the number on your account can effectively change if enough parties agree. You seem to agree with what I wrote so far...

An optimistic answer would be that blockchains are highly scalable, trustworthy systems of record that provide a number of benefits over a traditional database.

A pessimistic answer would be that blockchains are the latest buzz tech.

I like that dichotomous approach, but are those two in balance?

What makes it more scalable than a database when a characteristic is a redundancy of copies beyond what would reasonably be needed for sufficient backup? What makes it anymore trustworthy than a database with only trusted actors authorized with write access?

The trustful, or verifiable aspects of a blockchain come as a side-effect of independent actors competing for a financial reward. How does one replicate that dynamic in a single organization? Why would one bother trying?

a number of benefits over a traditional database

A blockchain doesn't provide a number of benefits over a traditional database.

A database isn't even a particularly fine-grained description of a blockchain.

A blockchain is a key component of an experimental currency inspired by a particular political ideology that is at odds with many aspects of western civilization, such as the benefit of trust in a society.

It's not really useful insofar as most people can tell in any other contexts. Even it's usefulness in it's native context is not highly established.

Why would a bank switch to blockchain internally? What's wrong with the existing ledger?

When two banks transfer money, there's no concrete thing being moved or transmitted. It pretty much just involves bank B telling bank A "You hereby have $100k of my cash". So A needs some kind of global ledger to keep track of how much cash B actually has, and double check whether it's already told C, D, E that they have all of it.

There's no reason in principle a blockchain is better at solving this problem than a trusted central bank. But if the existing central bank is moving slowly or has annoying policies, it's much easier to bootstrap a blockchain than bootstrap a new trusted authority.

So A needs some kind of global ledger to keep track of how much cash B actually has, and double check whether it's already told C, D, E that they have all of it.

https://en.m.wikipedia.org/wiki/Nostro_and_vostro_accounts

This is a long solved problem. No blockchains necessary.

As an aside this is a problem I often see in fintech pitches...

"When two banks transfer money, there's no concrete thing being moved or transmitted. It pretty much just involves bank B telling bank A "You hereby have $100k of my cash". So A needs some kind of global ledger to keep track of how much cash B actually has, and double check whether it's already told C, D, E that they have all of it."

No, there's no global ledger, and it isn't required. There are checks and balances and regulations that allow money to work in a completely distributed way. Unlike cryptocurrencies that use a centralized ledger (all the 'major' ones, at least).

I don't expect bank A to care about transactions between banks B and C, so the distributed blockchain doesn't sound very useful.

Maybe for more efficient netting? If A owes B $100 and B owes C $100 and C owes A $100 then no money has to ever change hands, even in the sense of crediting or debiting accounts at the central bank.

I wonder how many of these banks think they're using blockchain, and put out press releases stating that, because the engineers have lied to management in order to get projects underway.

Matt Levine has been speculating on this a lot recently: banks traditionally have hugely underinvested in infrastructure upgrades. I imagine if you're in IT at a bank, it must be awfully tempting to go "blockchain! secret project!" to get management on board because the suits are all caught up in the hype. Then you can just immediately pivot to Apache Cassandra (which is strictly superior to blockchain in pretty much every proposed business use of blockchain I've ever seen) and no one's the wiser.

This is exactly what is happening. "Blockchainism," a vague and shifting set of shared assumptions about voluntary custodial risk/reward sharing, vs "a blockchain", a particular encoding of a particular subset of these ideas into a white paper and corresponding client software, according to a particular framing of incentives motivating the people who created it, which may differ greatly from those who invest in it or speculate on it.

Blockchain is on its way to becoming a hollowed-out incantation, like "green technology" or "democratic" or "brawndo"

Would you please share references to your stated Cassandra superiority over Blockchain? Thanks.

Cassandra is real world proven tech with many actual, observable implementations "in the wild."

What even is a "blockchain," not in the context of something like bitcoin? How does a blockchain even work within a single organization? However it may or may not work, the dynamics involved would seem to make it qualitatively different from something like a bitcoin blockchain.

So you're asking for a comparison between something that does exist and something that doesn't. And in fact something that can never exist, if we are going to stick to the convention of treating names as more or less unique identifiers of a certain set of characteristics.

Try querying a blockchain.

Block explorers work great. Is that what you mean?

Reply on news.ycombinator.com