[SGVLUG] Bit order (was Linux based web-server appliance)

[David Lawyer]

On Sat, May 20, 2006 at 08:00:20AM -0700, Dustin Laurence wrote:
> On Sat, May 20, 2006 at 03:03:23AM -0700, David Lawyer wrote:
> 
> > On Fri, May 19, 2006 at 01:51:01PM -0700, Dustin Laurence wrote:
> > > Hmm.  The internet does no such thing; it's too high in the stack to
> > > know about bit transmission or, indeed, whether the transmission is
> > > serial or parallel.  Do you mean ethernet?  It certainly cares about
> > > byte-order, though, and specifies big-endian.
> > 
> > Doesn't the Internet have a physical layer ??
> 
> The answer is either "no" or "many, of all different types", depending
> on how you want to say it.
Correct.  Since I mentioned a physical layer, I meant it in the sense
of a physical layer consisting of various types of hardware.  Some
channels are carrying exclusive internet stuff, like between your
modem over the local loop of the telephone line to the local end office
of the phone Co.   There a codec converts it to digital and sends its
out over a cable where its time-division multiplexed with other
channels many of which are just digitalized voice telephone calls as
well as other modem calls.  For example T-1 (which is a big channel
containing 24 smaller channels.  As an aside, I checked on how T-1
does "modulation" and found it to be a lot more complex than described
in Tanenbaum's textbook "Computer Networks".

> The internet is named that for a reason; it goes across networks.
> Anything that carries IP is internet, no matter what layers below it
> are.  So it isn't even well defined to talk about the "internet's
> physical layer"; that isn't what the word means.

If there wasn't a physical layer, nothing could go over the internet.
We a just wasting time quibbling over semantics :-)

[snip]
> 
> > I just went on Google and typed "most significant bit first"
> > and "least significant bit first".  I got roughly 50k hits for each.
> > So it depends on the hardware.  For modems it's least first
> > (little endian) and until just now I incorrectly thought that the
> > whole internet would be like this, but it's not per the Google
> > results.
> 
> You assumed that something related to modems was predictive of real
> networking?!?

Not exactly, but I suspected it was true.  That's because the
Little-endian (bitwise) of modems was also used for dumb terminals and
they were the first "networks".  In the 1970s - 1980s there could be
perhaps 20-100 people connected to a single mainframe thru dumb
terminals.  They could chat and send email to each other.  It wasn't
really a network, since data was not sent in packets (with some
exceptions).  Then people started using modems and dumb terminals (or
emulated dumb terminals) to connect to computers (including bulletin
boards) over phone lines.  All of this was little-endian-bitwise.  So
I somehow thought that it would be likely that this data would remain
little-endian with other types of transmission such as ethernet, T-1,
and fiber, but it didn't turn out that way.

I just checked my dumb terminal with a volt-meter and it's
little-endian bitwise as I expected.  An oscilloscope wasn't needed to
look at the pulses since this terminal can be set to 2 baud (2 bits
per second) which can be read by the needle of a voltmeter.
Unfortunately, the modems I have don't support 2 baud :-).  What was
this slow speed used for?  Noisy lines?  With 10 bits/character
(includes a stop and start bit for each byte) it would take a few
hours to send one screenfull of text (about the size of a short
email).

			David Lawyer