News for hackers. Stuff that matters.
Re:On the dangers of a blockchain monoculture by geofft on Tuesday January 05, 2016 @02:12PM (#10845257)
It took me a while to understand why Certificate Transparency is sound without requiring proof-of-work, and Bitcoin is not, but the answer turns out to be fairly simple: you can merge two CT log entries without conflicts, but you can't merge two Bitcoin transactions. If I want to log "Certificate A exists for bitcoin.org" and you want to log "Certificate B exists for mtgox.com", both of our data items are valid on their own (and can be independently verified, just by checking the signature on the cert) and the ordering doesn't matter. So as soon as a write goes to some log server, as long as I trust the log server to propagate writes via the gossip protocol, I'm done; I don't need to care about merges or retries. (Also, it's possible for the log server to send me a signed promise to propagate writes, that I can later use to prove its misbehavior if that entry never shows up.)

On the other hand, if I have a transaction "Transfer 1 BTC to bitcoin.org" and you have a transaction "Transfer 1 BTC to mtgox.com", they don't necessarily commute or merge because of the double-spend problem: they might both be trying to spend the same 1 BTC.

There are a number of nifty decentralized systems you can build that don't have to think about double-spend problem, where all transactions commute/merge. Associating names with public keys seems to be one of them, as long as you allow a name to be validly bound to multiple keys. (A Keybase-style binding of external service usernames to public keys would permit this, since you simply have multiple valid Keybase proofs; the external usernames are arbitrated by the various centralized services. A Namecoin-style approach doesn't, since names are first-come-first-serve, and you don't want multiple people to be served the same name.)

There appear to be ways to make a money-transfer system avoid the double-spend problem: as I understand Stellar, it simply just makes the spending account legitimately owe both people money (and puts limits on how much each link can transfer at a time, so that the money is actually recoverable). But for non-money-transfer systems, if you can arrange your problem statement such that double-spending isn't a concern in the first place, you can use regular Merkle trees and avoid the entire proof-of-work overhead.

Re:On the dangers of a blockchain monoculture by murbard2 on Tuesday January 05, 2016 @02:53PM (#10845540)
Blockchains can count, and thus handle mutations. That's the key difference:

https://medium.com/@arthurb/what-do-blockchains-accomplish-e...

Re:On the dangers of a blockchain monoculture by geofft on Tuesday January 05, 2016 @03:13PM (#10845674)
Thanks, that post is very clear.