The essential use case of a blockchain is when you have something where you want an immutable history, and where, for whatever reason, it is either undesirable, or difficult to converge on a single entity to own that history. This arises in many, many places. Financial assets, source code, legal documents, etc..
It is not that you can't implement say, legal document version control without a blockchain. You absolutely can. But to do so, you have to trust a single entity to warehouse that truth for you. Blockchains are a way of solving that problem, just like Git was a way of solving it for open source.
Huh? "SQL would be centralized"
At no point is git required to use anything over a network or somehow is more decentralized than a database you pass around and insert data into.
A push request is a request to push the same data I pushed into my database into yours, whether its SQL or not makes no meaningful difference.
There are some interesting applications of blockchain in the examples you mention, but they are very case specific. For example, there would be resistance to creating a financial asset where trades can't be rolled back, for example (from governments and probably a number of traders). Legal firms might _like_ being able to destroy documents without leaving a permanent record. With source code we generally have highly trusted groups who are ideal central authorities (eg, I'm only going to get my kernel sources from kernel.org).
You're totally right. We're sort of mixing abstraction layers, here, of course you can implement Git on top of SQL and even SQL on top of Git! (since sql is just a query language + ACID constraints). But I suppose I was more responding to the spirit of the grandparent's statement, which was the traditional usage style of SQL. But you are absolutely correct.
> There are some interesting applications of blockchain in the examples you mention, but they are very case specific. For example, there would be resistance to creating a financial asset where trades can't be rolled back, for example (from governments and probably a number of traders).
Yes, that is certainly a possibility. However, if I were designing a financial system from scratch, I think i'd probably build in irreversibility at the protocol layer, and then allow institutions to sit on top of that protocol that implement reversible trades by delaying or abstracting settlement on that protocol. I.e. the current crypto model, where 99% of trading happens on exchanges, not on the blockchain. But, when you really do want an irreversible transaction, you bail out to the blockchain and you do it there. I think that is the most sensible model to build a modern, global financial system on.
> Legal firms might _like_ being able to destroy documents without leaving a permanent record. With source code we generally have highly trusted groups who are ideal central authorities (eg, I'm only going to get my kernel sources from kernel.org).
We do have highly trusted groups who are good central authorities, much of the time. But IMO that is something we've developed out of necessity. It's an artifact of our technical inability to collaborate with trust in the absence of intermediaries. What I think is so exciting about blockchains are their ability to alleviate that trust requirement in certain circumstances.
For instance, right now, financial institutions basically are in the business of trust. They do some other stuff too, but their course business is balancing risk and being trustworthy by having scale. JP Morgan isn't going to steal the money in your checking account, because if they steal your $10,000, they put their multi-billion dollar business at risk. So you can trust them not to steal your bank account balance.
But what systemic effect does that have? I would argue it has a centralizing effect, not just of trust, but of economics and innovation. If you have $10,000 to put in a savings account, you're not going to entrust it to a bank that only has $20,000 of deposits. Because they have a decent incentive to run away with it! This means that these sorts of institutions have very powerful incentives to grow ever larger, and be extremely risk averse. That is, they are incentivized not to innovate, because innovation is risky, and risk is trust-destructive.
Blockchains represent an opportunity to break these sorts of anti-competitive thickets, that rely on this intangible trust-asset to keep out competitors, in very much the same way that the internet helped say, blogs, compete with newspapers, by democratizing distribution. The internet democratized distribution, blockchains democratize trust.
What? Git doesn't solve that problem. Git allows peer to peer workflows, but on its own does not provide you with a network of forks or anything like that. For most projects, a single entity owning the source of truth is not only desirable, it's the current reality.
And before that, many projects got along fine with SVN and CVS, albiet with less fork friendly workflows.
I mean Git definitely enables better development workflows but I don't think it solves or attempts to solve the problem of having a source of truth.
I'll note that the three cases you mention aren't obvious candidates for blockchains either. Financial assets mostly work via central registries. Legal documents are generally kept by all parties, as each entity has its own perspective. Source code is usually centralized but sometimes perspective-driven. Blockchains are a possible way to solve some of those problems. But in the 10 years they've been out there, cryptocurrencies aside, I have yet to see a use that couldn't be done more effectively with a better-tested, better-understood technology.
The parent's point is that most of the time it's the other way around, with blockchains being hyped for use cases where they're a terrible fit.
Fossil is definitely not centralized. Fossil is implemented using SQL.
> The classic rebuttal against blockchains and other distributed technologies goes something like this: there's nothing a blockchain can do that a plain old SQL database can't do cheaper and more efficiently. That's true, but it's also missing the point, and I'll leave it at that.
That is definitely not missing the point when it is raised in response to things that are sold as "OMG blockchain!" but do not need (or often, even make use of) what makes blockchains unique. Which is, IMHO, at least 95% of the things that I come across marketed as involving blockchains.