pmetzger@lehman.com said:
Each DES block is eight bytes. You can't use hashing -- the idea is nonsense in context. Did you read the original post?
Yes, I did. If hashing doesn't work, you'll have to say why not. It's a technique that works in most other situations.
First of all, its actually twice that because you really need 16 bytes.
As I said, we can tackle this if anyone cares to...it's unclear that this is an invitation, but assuming it is: hashing gives a first-pass screening good for every 1/256 calculations, given the assumptions I stated. For each collision more work is needed...but you haven't invited that analysis, nor addressed it yourself.
Second of all, the method is still impractical. I was merely giving the easiest and most obvious attack on it.
Impractical? Your response to Karl implied that it was *impossible*. If you wish to apologize to Karl, and say that it is merely "impractical", then I will agree with you and drop the subject. The expense required definitely indicates that it is "impractical."
I see no reason to continue this. I don't think your argument has credibility.
Clearly you are preparing to drop the argument because you sense that your tactic of flaming didn't work. I welcome the lessening of flames, so thank you for that. We could use less flames here. Doug