> OpenSSL has yet another bignum implementation > separate from GMP. Any opinion on how they compare?
I am not very familiar with the openssl implementation. GMP should be faster, although I don't think the specific case of RSA on amd64 is one where the difference is the largest (but then, I haven't run any test, so don't trust me). The interface for GMP has its own bignum type, but it also has mpn_* functions that allow you to use them with your own bignum type (these functions expect buffers and sizes as arguments) whereas the openssl one seems to only work with its own BIGNUM struct. I remember some openssl people claiming that using their code was safer against some timing attacks, but I don't remember if this came directly from the bignum library or only the way it is used, so this might not be relevant. And obviously the license of openssl is more convenient for opensolaris than the LGPL. I know the paragraph above is not very helpful, but I thought I should still try to answer your question as well as I could... -- This message posted from opensolaris.org
