RIPEM is based on RSAREF. Has anyone on the list compared their own crypto
or a BSAFE-based verstion to an RSAREF-based version on the same machine?
For large messages, most of the compute-time is spent in the hashing and
symmetric-key encryption routines, and almost none of it doing RSA.
Running Cambridge PEM under a program profiler gives the following
percentage breakdown of CPU time. (To sign and encrypt a 200K message, using
asymmetric key management).
DES 67%
MD2 23%
RSA 1%
With the other 9% being all the other stuff, like reading and writing files
and parsing the PEM header.
These measurement were taken using a DES implementation that was not highly
optimised. They clearly show it's well worth spending the effort building
a decent DES implementation, and in fact that's just what we're doing.
Mike