3 ms·

A more useful contribution than this rant would have been a structured guide for how to do RSA without any of those issues. If they are all so well known, then,

A more useful contribution than this rant would have been a structured guide for how to do RSA without any of those issues. If they are all so well known, then, well, write down exactly how to do it properly and make a list of all attack classes together with a convincing argument why your guideline prevents them. But that would open up responsibility on the author's side, and who wants that? It's oh so much easier to just blast out a rant with some bits of information that are enough to illustrate your rant's point but far from sufficient to actually be helpful.

FartyMcFarter 6y ago
When it comes to security software, that doesn't work. There are so many ways to get things wrong, it's better to just use a well maintained existing library. Don't be the (next) person who makes one single mistake and throws everything away.
- ganafagol 6y ago
  Sure, I totally agree. But that was not my point. I was trying to convey that it's always easier to rant than to actually contribute to improving things. Mind you, the post does not just say "dont't roll your own". It essentially says "don't use RSA no matter which implementation". And the argument for that is not very strong. It's an arbitrary selection of what could go wrong. In which libraries does it? Maybe for the major implementations this is all just cold coffee and they don't have those issues. But then I go and use one of them, and since the author just wrote a random rant, they pull the next trick out the sleeve and claim the next thing this implementation got wrong. It's easy to be in that position. It's harder to compile a complete list so that I can go and see that my framework does not violate any of those issues, so it can be trusted just as much as his favorite ECC framework. But yeah that would not just take work but would risk being criticized for being incomplete, so just easier to publish a rant instead.

FartyMcFarter 6y ago

When it comes to security software, that doesn't work. There are so many ways to get things wrong, it's better to just use a well maintained existing library. Don't be the (next) person who makes one single mistake and throws everything away.

ganafagol 6y ago
Sure, I totally agree. But that was not my point. I was trying to convey that it's always easier to rant than to actually contribute to improving things. Mind you, the post does not just say "dont't roll your own". It essentially says "don't use RSA no matter which implementation". And the argument for that is not very strong. It's an arbitrary selection of what could go wrong. In which libraries does it? Maybe for the major implementations this is all just cold coffee and they don't have those issues. But then I go and use one of them, and since the author just wrote a random rant, they pull the next trick out the sleeve and claim the next thing this implementation got wrong. It's easy to be in that position. It's harder to compile a complete list so that I can go and see that my framework does not violate any of those issues, so it can be trusted just as much as his favorite ECC framework. But yeah that would not just take work but would risk being criticized for being incomplete, so just easier to publish a rant instead.