Details
Description
BouncyCastle's Cipher implementation strictly enforces IV specifications, while MRI does not and truncates to the proper length.
Attached failing test case which passes under MRI.
Full code sample listed here: http://www.headius.com/rubyspec/index.php/Openssl
Open Question: Should JRuby workaround this issue, should a bug be filed against Ruby, or is it between the BouncyCastle and SSLeay folks?
We'll leave it up to Ola. I don't see "more strictness" as something we should try to break, especially since as you say it could easily be reported as a bug against MRI or OpenSSL (not strict enough).