I'm not sure if bitcoind accepts blocks of the same block height twice (so 2 blocks with the same parent) if there is not a block following up.
A "Y" split might not be found (as your client already thinks it has the "correct" latest block), only if the block you happen to have is orphaned, and the other chain is forced upon you you keep the orphan data - if I remember correctly.
Ok I see, still it should be sufficient to detect most double spend attacks (assuming they are not targeted at a specific node). I guess this is something i could patch in bitcoind, I need to read more into how blocks are advertised.
Very nice! How are you calculating the cost to mount a 50% attack? Am I correct in reading that you estimate it to be $12.7 billion? That's far greater than what I've estimated in the past. That would be awesome if accurate, but I think that's probably way too high.
Should be 12.7 million.