[7122] in Kerberos
More errata for RFC 1510 ?
daemon@ATHENA.MIT.EDU (Darren Reed)
Thu Apr 18 19:15:31 1996
Date: Fri, 19 Apr 1996 08:59:18 +1000
From: darrenr@nabaus.com.au (Darren Reed)
To: kerberos@MIT.EDU
In the process of working with a 3rd party software company, implementing
portions of RFC 1510, we've noticed a previously undocumented discrepencies
with the RFC and real-life Kerberos (yes, we checked the errata). It would
appear that although the RFC states that the CRC is NOT to be calculated
over the data, including the padding, writing the data out to a file from
the CRC routines shows that the data being CRC'd is padded and that the
length is a multiple of 8 bytes. This disagrees with the 2nd paragraph on
page 70.
Also, it would appear that the CRC is reversed before storing. This looks
like a hangover from little endian machines, however, I suspect that any
change to this will break Kerberos interoperability. The value returned by
the CRC-32 routine in Kerberos5 Beta5 will not match the returned value for
a normal CRC-32 over the same data.
I've not yet looked at either MD4 or MD5...
Darren
