Note that specific hand-picked curves can be much faster than random curves (like the Brainpool ones). Crypto++'s general elliptic curve code isn't terribly efficient, which makes this worse.
According to Tanja Lange's earlier post, NACL will have digitial signatures, presumably on a fast curve, and I plan to implement it too in Crypto++ after they release the code or specification. So if speed is a big concern, you might want to either wait for that, or plan an upgrade path. It should be around 5-10 times faster than the current ECDSA implementation in Crypto++ with Brainpool curve. _______________________________________________ tahoe-dev mailing list [email protected] http://allmydata.org/cgi-bin/mailman/listinfo/tahoe-dev
