Anyone with a bit of a deeper network and cryptography knowledge care to comment on this technical proposal:
https://bitcoinmagazine.com/articles/bitcoin-ng-or-how-cornell-researchers-think-a-radical-redesign-can-solve-bitcoin-s-scaling-issues-1447108649Basic idea seems to be to decouple proof of work blocks from transactions blocks (while keeping the two connected, obviously), without the usual trade off that means more tx -> bigger blocks.
Looks plausible to me, and actually a lot less "radical" than, say, turning Bitcoin into a proof-of-stake system. That said, I don't have sufficient technical knowledge to judge if it only appears plausible at a glance or if there's a catch I don't see right now.
(... he asked on the Wall Observer thread, thinking "What could possibly go wrong?")
This is an idiots (that is me) take on it:
It addresses mining centralization. But mining centralization is sort of a lost cause because of the economics of mining. And sort of not a problem because of the economics of it. Mining will go the areas with the cheapest electricity, globally. Sending work to a server connected to good fiber can be done anywhere in the world. There is a built in incentive to keep the network robust and decentralized to a certain degree. This has been put to the test recently (ghash.io) and I'd say Bitcoin passed.
The idea of "poison transactions" is a bit unclear to me. If it solely relies on lost rewards as a way of incentivizing miners to remain honest, then I think that is sort of naive. If anyone wants to build a sufficiently large farm to mine blocks on their own, then they either want to protect the network or attack the network. Not commit theft.
As far as I can see Bitcoin NG doesn't address node centralization, which I think is the most contentious issue. All of these transactions on the microblock will need to be communicated. And I don't see how unlimited microblocks every 10s will lead to less bloat or less network demands on the node.
Edit: Turns out
Gavin Andresen agrees with me on the last part:
"It's an interesting idea, but it isn't a scaling solution-- the same amount of transaction data has to get to every fully-validating node, it doesn't matter if it is sent in the microblocks of the Bitcoin-NG proposal or the blocks we have now."