Re: [Cfrg] scrypt password-based key derivation function

[Joachim Strömbergson <[e-mail]>]

Aloha!

On 2012-09-24 11:55 , Simon Josefsson wrote:
> Colin and I have published a draft describing Colin's scrypt key
> derivation function.  We would appreciate review of the document:
>
> http://tools.ietf.org/html/draft-josefsson-scrypt-kdf

Great initiative and good work! I'm trying to follow the draft and build 
an implementation just to see if I can understand it. One thing that 
caught me is the test vectors for Salsa20 in chapter 7.

First off, Sals20 as defined on the DJB page you point to 
(http://cr.yp.to/salsa20.html) works in arrays of 16 uint32_t values. 
The vectors you provide are in byte form. I'm probably thick, but for me 
it is not 100% non ambigious how to combine the bytes into the correct 
sequence of words.

You could also mabye add another set of vectors where all but one byte 
is zero, alternatively that the first 32 bits are all set and the rest 
is all zero. This would help out debugging the endianess/ordering.

-- 
Med vänlig hälsning, Yours

Joachim Strömbergson - Alltid i harmonisk svängning.
========================================================================