[2580] in Kerberos_V5_Development
Re: Prototype hell
daemon@ATHENA.MIT.EDU (Theodore Y. Ts'o)
Fri Oct 17 16:25:53 1997
Date: Fri, 17 Oct 1997 16:25:18 -0400
From: "Theodore Y. Ts'o" <tytso@MIT.EDU>
To: Ken Hornstein <kenh@cmf.nrl.navy.mil>
Cc: Assar Westerlund <assar@sics.se>, krbdev@MIT.EDU
In-Reply-To: Ken Hornstein's message of Thu, 16 Oct 1997 09:24:35 -0400,
<199710161324.JAA12465@ginger.cmf.nrl.navy.mil>
Date: Thu, 16 Oct 1997 09:24:35 -0400
From: Ken Hornstein <kenh@cmf.nrl.navy.mil>
>What SHA code are you using? Most of the MD{4,5},SHA implementations
>that I've seen assume that there's a type which is exactly 32 bits
>wide and break on Crays. We had to fix that as well in our code.
Ummm, I dunno where the SHA code in V5 came from .... Ted?
I realized the attribution of the code had gotten chopped out somehow.
I'm not sure why; I wouldn't have done that deliberately. A quick
comparison with a collection of SHA implementations that I have looks
like it was derived from the public domain one done by Peter Gutman.
A quick check looks like the 64-bit portability problems lies in two
areas. The first is in the 32-bit rotate left instruction, which needs
to have some AND instructions to mask off any extra bits in a 64-bit
architecture. The longReverse function also makes some really bad
assumptions about the size of a long.
- Ted
