Comment by Lazare | Hacker News Reader

Comment by Lazare | Hacker News Reader

Parent Full thread

> It takes some financial assets 3-5 days to settle after the initial transaction goes through, weaving its way through many database systems (some many decades old) and middlemen who each take a cut, e.g. clearing houses. Any established industry has lots of cruft that builds up.

Yes, but centralized databases are a solved problem. It doesn't take 3-5 days for Amazon to process an order, or for github to process a pull request. The problem is here:

> weaving its way through many database systems (some many decades old)

Right, so you could replace the entire thing with a single centralized database, and cut the processing time to ~0. And you wouldn't need to develop any new technology to do it.

> For this reason Santander estimates that banks can save $15-20bn per year collectively by using blockchain technologies to make their infrastructure more efficient

Assuming that is correct, that implies you could save even more money by using existing database technology. Right? What I'm looking for isn't an argument about why banks need to update their technology (they do), or why the existing system is almost unbelievably archaic and inefficient (it is).

What I'd like to know is why blockchains are the correct upgrade path when they appear to be strictly inferior to other existing technologies along the dimensions that banks are currently struggling with.

Of all the places blockchains make sense, an industry with universally used, completely trusted, and heavily regulated central clearing houses seems like the worst possible case. No?

True, the KYC makes slightly more sense. Even there I have to wonder though. If I was hired as the CTO for an industry consortium to design a secure way of allowing customers to store, manage, and share their PII, I can see a lot of potential problems, but none of them feel like something a blockchain would help solve.

Or to put it another way: How come Amazon never talks about using private blockchains to manage inventory or payments to their sellers? Wouldn't all the arguments in favor of blockchains among banks apply even more strongly to using them within Amazon?

Great questions.

It comes down to speed. We expect money to move as fast as information, but instead it moves like products.

In meat space it's OK to work with batched 30-60 day pay cycles.

But finance needs to move faster.

The complexity and cost is in verification. Electronic records have to be verified against paper records which have to be verified against records in other systems, which have to be compared between institutions. Additionally there are many instruments with varying characteristics, and all of it has to flow in real time.

It might be unsolvable without technological innovation like blockchain.

> It comes down to speed. We expect money to move as fast as information

All current implementations of blockchain are inherently slower than more traditional approaches.

That is partially because of the unpermissioned nature of the consensus mechanism. With a closed system, you don't need to worry about such issues. Banks mostly trust each other.

So why do they need a blockchain then?

> Yes, but centralized databases are a solved problem. It doesn't take 3-5 days for Amazon to process an order, or for github to process a pull request. The problem is here:

With the current state of the art[1], it takes several days to settle a transaction. In 2016, this is really not ideal.

> Right, so you could replace the entire thing with a single centralized database, and cut the processing time to ~0. And you wouldn't need to develop any new technology to do it.

I could think of several problems, mainly around trust. Who hosts it? How is access managed? A regular SQL database makes sense if it's within a single institution, because presumably you trust your own employees to enter trade data with the best of intentions, but this trust isn't there for external parties.

It's not the storage which is the issue, it's the verification. If you are (made up example) JP Morgan, and you've sold $500m in bonds to a mutual fund, both parties know for sure that everything went smoothly if their copy of the blockchain tells them that it went ahead (just like you know that your Bitcoin wallet contains x bitcoin, and if you send that balance to an address, that transaction is absolutely confirmed beyond doubt). The transaction and settlement happen at the same time, instead of a 3 day process which is the situation today (for stocks). They don't need an expensive, external authority to tell them that everything went smoothly, they can confirm it themselves much faster and cost effectively.

Here's a presentation by UBS, exploring possible use cases for what they call "smart bonds": https://www.youtube.com/watch?v=jgJxEHwj-XU (around 4:00 he speaks about the benefits of smart bonds)

> Of all the places blockchains make sense, an industry with universally used, completely trusted, and heavily regulated central clearing houses seems like the worst possible case. No?

I could be wrong, but I think what's being proposed with these private blockchains is that the central clearing houses' role would be reduced to being an arbitrator in the case of disputes between industry players, e.g. in a proof of stake network, the own enough % to confirm a transaction or roll it back (maybe their role is even eliminated entirely). Whether or not this actually happens will depend on legislation of course, not technology.

> Or to put it another way: How come Amazon never talks about using private blockchains to manage inventory or payments to their sellers? Wouldn't all the arguments in favor of blockchains among banks apply even more strongly to using them within Amazon?

Probably because Amazon is a hugely powerful entity, and it's sellers have no power to demand equal access to a shared database. All (most of?) the large, important transactions happen on an Amazon database, which isn't the case in the financial sector. Plus Amazon is a growing business, whereas in investment banking growing revenue is a huge struggle post-2008 with newer regulations so there is a rush to cut costs and inefficiencies.

Disclaimer: I'm not an expert, this is based on what others have told me, and research papers I've read.

[1] http://www.investopedia.com/terms/s/settlement_period.asp

> I could think of several problems, mainly around trust.

And this is the problem that blockchain+proof-of-work solves so well.

Financial companies are however different beasts. They are not and do not want to be anonymous. While they don't fully trust each other, they do have significant trust in each other and they are typically regulated by a central authority. The blockchain-based solutions advocated by R3 seem to have a number of services that must be trusted by all participants, including time stamping.

Traditionally when coordination has been required between the banks (e.g. fast clearing solutions for retail banks), a new company is created that is joint owned by the members, and is therefore trusted enough by the members to run the centralized database.

To avoid the single-point of failure aspect you can run a distributed database across multiple nodes.

I still haven't heard a good explanation of why solutions like those aren't adequate.

> It's not the storage which is the issue, it's the verification. If you are (made up example) JP Morgan, and you've sold $500m in bonds to a mutual fund, both parties know for sure that everything went smoothly if their copy of the blockchain tells them that it went ahead (just like you know that your Bitcoin wallet contains x bitcoin, and if you send that balance to an address, that transaction is absolutely confirmed beyond doubt). The transaction and settlement happen at the same time, instead of a 3 day process which is the situation today (for stocks). They don't need an expensive, external authority to tell them that everything went smoothly, they can confirm it themselves much faster and cost effectively.

That isn't what clearinghouses do in a legal/functional sense for participants.

Reply on news.ycombinator.com