Date: Mon, 18 Feb 2013 18:36:45 -0800
From: "Dennis E. Hamilton" <dennis.hamilton@....org>
To: <discussions@...sword-hashing.net>
Subject: RE: [PHC] Coding of the in[inlen] array for PHS( )

Now I am more puzzled than I am by Larry's comment about presumed order.

I am going to risk over-stating the obvious just to make certain that I, as
a newcomer, am on the same page about what the tacit understanding is
concerning what PHS( ) effectively operates on.

 - Dennis

TL;DR

 1. Let me look at it another way.  I presume that the each of the byte
arrays (in[inlen], out[outlen], etc. can be viewed as a canonical string of
bits, e.g, in[0]||in[1]||...||in[inlen] where each byte 
has its bits laid into the string as b[7]||b[6]||...||b[0], exactly as the
bits weigh into interpretation of the byte as an unsigned integer, are
expressed in hexadecimal, etc.  The expression of in[] in hexadecimal is
simply the concatenation of the hexadecimal digits of the bytes, in their
standard order.  (If one were operating on strings of bits whose length is
not a multiple of eight, it would be left-justified in the byte array and
any padding would start in the unused low-order positions of the last byte.)
  The hexadecimal-notation expression of test vectors and the expected
results conforms to this abstraction.

 2. The algorithm of a PHS( ) can be defined abstractly as operating on such
structures, however organized into blocks, padded, and computed over, with
the results of the same form.   When operating on wider units, there may be
collection into binary words, use of (modular) arithmetic and logical
operations, etc.  Those operations will assume a particular way of chunking
and ordering the material in widths larger than one byte.  That's how I
understand the algorithmic specifications of the NIST FIPS 180-n Secure
Hashing functions.

 3. Any differences in how a particular platform provides its way of
chunking and ordering and operating on various word-sized chunks should not
be visible to any user of the API -- that is, the order at the API is
sustained and the result is indistinguishable from it being carried out via
(2) regardless of what is actually done to accomplish that.  That's what I
see various implementations of NIST FIPS 180-n Secure Hashing Functions
accomplishing, and their APIs are consistent with that.

Is what is being said here tantamount to the same thing, but not so
long-winded?  

PS: I agree that the size_t, int, and unsigned int values represent numbers
in exactly the way that is done on the platform running the PHS( ) code and
their memory structure is not of concern in the implementation of PHS( ).  

PPS: I'm assuming serialization/deserialization of the cost parameters in
input-output is an outer-protocol consideration (likewise for the password,
salt, and hash) but the reference implementation, PHS( ) would not reflect
any of that.  I am also using the ordering of in[0],in[1],...,
in[i],in[++i],...,in[inlen-1] (likewise for the salt and hash byte arrays)
as being the canonical ordering, regardless of how that is achieved in a
particular processor's storage organization.



-----Original Message-----
From: Solar Designer [mailto:solar@...nwall.com] 
Sent: Monday, February 18, 2013 16:54
To: discussions@...sword-hashing.net
Subject: Re: [PHC] Coding of the in[inlen] array for PHS( )

On Mon, Feb 18, 2013 at 03:57:21PM -0800, Larry Bugbee wrote:
> On Feb 18, 2013, at 3:00 PM, Solar Designer <solar@...nwall.com> wrote:
> > In PHS(), we deal with bytes.
> 
> +1
> 
> ...and said bytes/byte strings/byte arrays should be specified to be in
network byte order.

No, byte order does not make sense for the password and salt inputs, and
the numeric inputs are just numbers (in the PHS() interface, they're
just size_t and unsigned int variables that have the correct values in
the host's native representation).

Alexander

Powered by blists - more mailing lists