On Tue, Sep 18, 2012 at 09:16:23PM +0200, Simon Josefsson wrote: > Nick Galbreath <[email protected]> writes: > > * need reference to base64, most likely http://tools.ietf.org/html/rfc3548 > > As I learned the hard way there several, and crypt-256/512 uses a > > very custom one. > > RFC 4648 even. :-)
Actually, neither. I think we should in fact reuse what Nick called a "very custom" base 64 encoding, because it is what has been traditionally used by all crypt(3) flavors except for bcrypt since 1970s. It happens to be simpler and slightly shorter than base64 (no padding). Alexander
