Does the number of security vulnerbilities scale linearly or exponentially with the number of threads?
But seriously, why would you make something like this in assembly? I can understand not wanting to use libraries, but even using C would make it portable.
I would personally prefer to not to use assembly & C for handling untrusted input, which is something that a web server does constantly.
This still seems interesting, but I can't understand why assembly was chosen for such a complex project. Compilers are pretty good at optimizing & computers are very fast, so I don't see there to be anything to gain by doing this.
They say that their cryptography implementations are fast, but that's not the only thing that's important.
Maybe, but it just seems a bit too big for that to me.