Octets calculation?

David Bolen db3l at fitlinxx.com
Thu Jun 12 02:13:04 CEST 2003


William Yeo <wcyeo at shaw.ca> writes:

> In article <0modevgetee4eh0pg2k7s4qhacq7s6bbkm at 4ax.com>,
>  Tim Roberts <timr at probo.com> wrote:
> 
> > "Greg Ewing (using news.cis.dfn.de)" <g2h5dqi002 at sneakemail.com> wrote:
> > 
> > >Tyreal wrote:
> [snipt]
> > >
> > >In my experience, this supposed ambiguity about the meaning
> > >of the word "byte" is something that only exists in the
> > >rarified imaginations of standards writers. In the real
> > >world, nobody ever uses it to mean anything other than
> > >8 bits.
> > 
> > Today this is true, but remember that POP, like most of the Internet
> > protocols, is very old (in computer years).  At the time the early Internet
> > RFCs were written, there were a number of popular mainframe computers that
> > did not use 8-bit character sets.  The Control Data 6000s and Cybers used
> > 6-bit characters.
> 
> 
> Ah yes, and remember Honeywell in the glorious 70's whose systems used 
> 36 bits arranged as 6 6-bit bytes or 4 9-bit bytes depending on whether 
> you're in batch or timesharing mode.

Not to mention the DEC PDP-10 series (also 36-bit machines), which
were certainly popular at several of the larger universities that had
significant impact on the early Internet protocol development.

On those machines, the "byte" instructions actually operated on
"bytes" of specified bit length, so you could pack all sorts of
different sized bytes into a 36-bit word.  I believe the processor
manual defined a "byte" as any contiguous set of bits within a word
(36-bit).

How relevant bytes versus octets may be in today's typical hardware is
debatable, but it doesn't seem harmful to me to have the more precise
nomenclature used in the standards.

-- David




More information about the Python-list mailing list