Re: Cohen's paper on byte order

From: Douglas A. Gwyn (DAGwyn_at_null.net)
Date: 05/01/03 

Date: Thu, 01 May 2003 01:48:04 -0400

Mok-Kong Shen wrote:
> One point is, what you 'repeatedly' seemed to have
> ignored, that such an octet-stream interface problem
> affects 'also' other computer applications (crypto or
> not) and certainly not sigularly AES. Hence an amendment
> of the AES document for that reason isn't appropriate
> in my conviction.

The AES document has a particular characteristic that
causes a significant problem in interfacing to the
mainstream computer world: it specifies I/O only in
terms of a pure bit stream.

> Now, aren't the UNIX files byte (octet) streams?

Not *necessarily* octets, although I'm not aware of
any Unix-based systems *currently* being marketed
where I/O is performed using units other than octets.

> How is 'there' the 'interface' problem treated?

There is no interface problem, because ...

> E-mails and Webs are all character (octet) streams.

... octet streams are so universally used that we have
in place sufficient standards to ensure that octet
values are not corrupted whe data is exchanged among
computer hosts of all kinds.

> ... I think that this fact indicates that
> most hardware have bits within octet in big-endian
> (as the URL I cited says), hence there is no
> conformity problems in the majority of pracitcal
> cases. For the presumably comparatively seldom cases
> of little-endian within octet, conversions would
> surely need to be done for files coming from or sent
> to big-endian machines.

WRONG!!! Go study the explanations!!

> ... I started a thread
> in comp.arch entitled 'Q: Ordering of bytes and bits'.
> In my understanding, people there don't quite agree
> with you. So you may like to join that thread.

No thanks! I hope that you did not try to present
*any* argument as representing my position, not even
by an out-of-contxt cut-and-paste. If so, please
post a follow-up stating that whatever you posted
most assuredly should not be taken as an accurate
representation of my point of view, but constitutes
only your own attempt to present *an* argument,
which might bear no relation to my own presentation
of the same issue within the relevant context.

Relevant Pages

  • Re: Cohens paper on byte order
    ... But the bits in an AES byte are not 'related' ... That binary integer representing the octet ... AES document the bits of a byte are labelled by 0-7. ... trivial correspondence between the bits of an octet ...
    (sci.crypt)
  • Re: Cohens paper on byte order
    ... In the other case, where the octet has ... most recent follow-up to Bryan Olson). ... there is a natural and trivial mapping of between the ... there isn't any problem with the current AES document ...
    (sci.crypt)