| HN Mirror

A blockchain is a modification of a hash chain to make it usable in cases where different transactions conflict with each other. In the case of TLS certificate transparency, Go packages, etc., a log entry "This object with this hash was published" does not conflict with a log entry "That object with that hash was published." You can append them to the chain in either order and it's fine. For a financial ledger, "This money was moved from A to B" conflicts with "This money was moved from A to C," and so one needs to be applied first so that the other can be rejected. If both transactions are signed it doesn't matter which one (for the correctness of the system, at least), but all users have to come to consensus about which of the two is accepted. Otherwise you have the "double-spend" problem.

Bitcoin's blockchain solves this problem with "proof-of-work," i.e., making it computationally difficult to add another entry to the hash chain. Because it's difficult, it's unlikely that both transactions will be accepted at the same time, which serves as a publicly-verifiable way to arbitrarily pick one. There is a small financial incentive for the network entity that successfully solved the computational problem, hopefully offsetting their expense in "mining" the "block." (As a further means of making sure only one is accepted, even if both are mined in quick succession, it's common practice in Bitcoin to wait for a few more blocks to be mined before B or C accept the transaction, which forces the network to accept one chain or the other.)

Satoshi's innovation was that proof-of-work can be used to allow arbitrary / anonymous entities to participate in this system without risking someone creating several thousand anonymous entities voting in their favor, or creating one view of the hash chain for B and another for C to convince them both to accept the transaction. (Satoshi's innovation was not the append-only cryptographically-verifiable store; that's just a hash chain. It was securing updates to that store without a central trusted coordinator.)

There are other ways to arbitrarily pick one transaction or the other, e.g., using a distributed consensus algorithm and admitting participants based on "proof of stake." Or you can redefine the problem to allow double-spend to a limited extent and make A responsible for paying out both B and C - this is roughly Stellar's approach.

But if you don't have a double-spend problem (or if you have a central trusted coordinator), you just need a boring old hash chain, which can be implemented much more efficiently - and possibly expanded to a hash tree, as described in this article.