But then can't the malicious 51% ignore these votes (they are transactions, right?)?
They can ignore them, but they will be stuck with 51 delegates while the honest chain with the majority of shareholder approval will move up to 100% participation and eventually be longer.
Exactly what I was getting at.
This is also the reason why I said you can only "take over" the network with 100% delegates.
Even an honest 49% can still recover.
This is perhaps a topic for a future blog post. You can "recover" from any attack as long as the attackers never gain more than the long-term average delegate participation rate. The reality is that you would recover with a developer released hard-fork / check point.
Only an idiot would trust "the longest chain" metric on its own. The longest chain is just a "short term" consensus building approach, it is not the basis of long-term consensus.
Right, so a 51% attack is not viable in the long term.
A hardfork/checkpoint is more or less "cheating" (requires a human factor) but still a viable solution.
It seems that as more malicious delegates are voted in (or just turn malicious), it's harder for the network to "naturally" recover, but it
can still easily recover from a 51% attack.
A lower % of an honest network can still recover, and that's where Bitshares wins with DPoS.
With Bitcoin, a "Longest chain" isn't really longer than any other chain, but the difficulty is higher.
With Bitshares, the "Longest Chain" is asserted because any other (malicious) chain has fewer blocks, because malicious delegates ignore honest blocks that vote them out.
A malicious delegate can't just vote in another malicious delegate because she needs enough stake in the chain for that to happen.
Having stake in a malicious chain goes against the very premise of "zero stake".
Man, I love Bitshares