Can you show what exactly you were doing to "install"
BouncyCastle? Just adding the jar to your classpath
isn't enough, you have to register the provider.

Regards, Lothar

Last time I read anything about it, about ten years ago, an important number
from Fermat was factored. It's a one with about 512 zeros after it, plus one
(or something like that). He said (or one of his four rules determined) that
it was not prime. Then in about 2000, probably before, someone finally
factored it. The writers of the time figured that a lot of human input went
into the feat, because it had so many zeros in it. Still, if you think about
it, the task ran for about thirty years, because Fermat was long dead, but
recognized as a major contributor to PGP. Every additional bit doubles the
time (less than that, but not much). Add another 512 bits, and it still
can't be done by brute force, plus you will get major advantages from
employing expensive decrypto mathematicians who make these spectacular and
baseless but effective hunch moves to eliminate like five hundred million
instructions per hour. I am sure there are lot of crypto-wizards who are
saying that our computing power doubled five hundred times since 2000....but
it did not. Moore's law is losing momentum. I spend a whole lot more time
being wrong than anything else in crypto. Factoring in linear time
proportional to bits in a key would be astonishing to a lot of people. 1024
is the standard. Zimmerman was recommending it in 1996.

I saw that a recommendation in an RFC that symmetric keys be exchanged
periodically. There is no point to that, because symmetric keys are tighter
than public keys. The older a symmetric key, the better.