This didn't happen on BTER because they wait for 100 confirmations before any deposit.
This is not true when the attacker has 51% of the resource (either computing power for PoW or stake for PoS). With >50% of the critical resource it is a mathematical certainty the attacker will eventually build the longest chain. 10 confirms, 100 confirms, 20,000 confirms isn't sufficient to guarantee the attack can't double spend.
The attacker simply picked the weaker of the two exchanges however if both mintpal and bter used 100 confirmations the attacker could have double spend both by building an parallel chain containing the double spends that resulting in a 101+ block reorg.
Whilst it is a mathematical certanity that an attacker would build the longest chain, it is not a mathematical certanity that the attacker would be able to build a chain of required length.
Probabilisticly, if you consider a case where our would be attack holds 80% of coin age at any give point in time and if we assume that this holding of coin age does not dimminish after he finds a block, then the probability of the attacker creating a blockchain in which he can take the piss is;
proability of finding next block aprox 80% = 0.8
proability of creating block chain of length n = 0.8^n
ie. the probability of attacker creating a block chain of length 3 to deposit on mintpal is
P(attack chain length 3) = 0.8^3 = 0.512 or 51%
so if we observe 100 individual 3 block creations where our attacker holds 80% of coin age 100 times we would expect the attacker to be able to create a 3 chain block 51 times out of 100.
ergo, mintpal is fucking amateur hour.
lets compair the case with bter.
chain length require is now 100, so n = 100
P(attack chain length 100) = 0.8^100 = 2.0 x 10^(-10) = 0.00000000002
so we would expect our attacker to be able to create a blockchain capable of double spending bter aproximetly 1 in every 10 billion instences of a 100 block long chain being created.
Of course if the attacker has more than 80% total coin age this again increases, however, I must point out than once coinage has been used to secure 1 block, it is nulled, and so it is not accurate to assume that the probability is discrete.
More likley the same calculation would look something like this for whatever blocks....
P(attack chain of length whatever) = 0.8x0.79x0.78 etc etc etc.
Clearly,there will be no significant difference for a attack chain of length 3, but there will for an attack chain of length 100 (but it is already very, very small).
tl;dr, yes, but mostly hell no, more confirms make a HUGE difference, mintpal fucked up.
-----------------------------------------------------------
edit* also, this would be nulled if the fucking exchange staked.
