Re: Hmm Test Vectors....

From: Mark Wooding (mdw_at_nsict.org)
Date: 06/29/03 

Date: 29 Jun 2003 15:23:43 GMT

Michael Amling <nospam@nospam.com> wrote:

> I don't think you can pin this on EBCDIC. People who made the
> decision about parity would no more use EBCDIC characters as a DES key
> than use ASCII characters as a DES key. With only 56 bits, it's
> important that the key be chosen uniformly, not limited to printable
> characters.

That said, I seem to recall that IBM's RACF system used (still use?) DES
to hash passwords by encrypting the user's id (textual login name) with
the password as a key -- all using EBCDIC. My recollection doesn't go
as far as explaining what happened with the password bits which are used
for parity checking in a DES key -- obviously the parity bits are in the
least significant bits of each octet unless IBM's done something really
weird.

(If someone could tell me I've made this up or fill in the gaps in my
memory, that'd be great.)

-- [mdw]