Newsgroups: comp.arch
From: hoey@zogwarg.etl.army.mil (Dan Hoey)
Date: 24 Mar 92 22:36:25 GMT
Subject: Re: Standard term for bit ordering

bran...@inf.ethz.ch (Marc Brandis) writes:

>I ran into the problem of finding an appropriate term for the bit ordering
>of different architectures. While big-endian and little-endian are common
>terms to refer to different byte-orderings, I am not sure on whether this is
>also widely understood for bit-orderings.

As I recall the defining document--``On Holy Wars and a Plea for
Peace'' or something like that--the terms were most emphatically used
to refer both to bit orderings and to byte orderings, as well as to
orderings of any other addressable unit.  In the case of some
machines, the bit-addressability is not obvious, but there is usually
some instruction or another that associates numbers with bit
positions.

And I would not completely discount the ordering imposed by the plans
and documentation.  If they impose a numbering scheme, that is an
endianness, and it may or may not agree with the endianness implicit
in various architectural features.  It is possible to write accurate
documentation whose endianness conflicts with the architecture, it
just is messier and harder to understand.

The terms ``big-endian'' and ``little-endian'' are an extension of the
more traditional terms ``MSB-first'' and ``LSB-first''.  ``First''
usually refers to the order imposed by the addressing system in use,
though it often referred to the time order of some bit-serial
transmission technique.

Dan Hoey
Hoey@AIC.NRL.Navy.Mil