The author attempted to solve issue #1 by adding the 1000-msec delay for the first challenge and unrelated-password request throttling. That solves the issue for attackers using the client API (they can't narrow their search space).
The author has already acknowledged in the article that, were an attacker to get ahold of the stored password hash, they would have "everything needed to carry out brute-forcing."
I agree this is probably a deal breaker and defeats the purpose of using a hash function with a work factor like bcrypt.
The author has already acknowledged in the article that, were an attacker to get ahold of the stored password hash, they would have "everything needed to carry out brute-forcing."
I agree this is probably a deal breaker and defeats the purpose of using a hash function with a work factor like bcrypt.