If an attacker steals a DB of passwords, it is only a matter of time before computing power catches up and is able to crack the list, regardless of the methods used to hash the values. Advances in cryptography are rare, advances in processor capabilities are not. bCrypt may be the best we can do at this point to delay this inevitability, but programmer's shouldn't come away from this thinking that using bCrypt "solves" this problem. An interesting question related to this is how long a time period is considered "safe" enough to protect a stolen list? If the list is protected from brute-force cracking for 3 years after theft, is that enough time to render the passwords unusable? 5 years? It seems like the answer to this question would be used to calculate an appropriate value for bCrypt "speed".