You're right that I don't have a very deep cryptographic background, but from everything I've seen and read, dictionary attacks still work on bcrypt'd passwords, just more slowly than otherwise. This is the part I want to know if I'm wrong about; is there some a reason a dictionary attack doesn't work on bcrypt? I understand that it will take longer to get 100% password recovery out of a bcrypt'd database than a database hashed with SHA-256 or SHA-512. Let's say that's not relevant because our theoretical attacker has A) a computer from the future which is fast enough to blow through bcrypt just as quickly as SHA-* or MD5; or B) is only interested in one user's password, so the expense is bearable. In these situations, does bcrypt provide extra protection that isn't routinely implemented in a standard hash?
My understanding is that bcrypt does not provide much besides the slowdown. I'm not trivializing the slowdown, I understand completely why it is beneficial to keep passwords away from bad guys as long as possible after a data leak. I am merely saying that if I'm correct and the only thing standing between bcrypt and SHA-* for a captured password database is X iterations of silicon, I don't really see using SHA-* hashes as "completely irresponsible".
I would have ignored your original comment and avoided pedantry except for your original assertion that your salting did something to mitigate the risk of not using bcrypt. It does no such thing. There's nothing else for us to argue about.
First, they had access to the complete Gawker database, could have downloaded it and run their attacks on their side. They still would have needed to use blowfish/bcrypt (afaik), but your high rounds are useless if an attacker gets the actual data out of your infrastructure, just like they are with normaler hashes.
Second, the Gawker guys had no salting, and they used DES, which, as you know, was deprecated before many HN users were born. This made cracking the passwords really easy ... not that it would be that much harder with a conventional salting mechanism or even with bcrypt.
I was not contending that bcrypt is not like really great, just that it's not "completely irresponsible" to not use bcrypt or scrypt. bcrypt is still vulnerable to dictionary attacks, the only difference is the amount of time it will take to get everyone's password out, which, again, is useless in a case like gawker where the dumps are distributed all over the internet for anyone with an interest to peruse.
Please correct me if I'm wrong.