[ih] ARPA initial IMP-IMP line speed

Craig Milo Rogers rogers at ISI.EDU
Sun Feb 22 23:02:43 PST 2004


On 04.02.23, David L. Mills wrote:
> Before 1968 the higest speed you could get on an ordinary telephone line
> was 300 bps full-duplex (Bell 103 FSK), 1200 bps half-duplex (Bell 202
> FSK) and 2000 bps half-duplex (Bell 201 DQPSK). The last could go at a
> blinding 2400 bps full-duplex on a 3002 leased circuit. You couldn't get
> anything higher on an ordinary phone or leased line until the middle
> eighties.

<begin tangential reminiscence>

	By the mid-1970's you could get a 1200 bps modem with a 150
bps back channel; the combination provided asymmetric full-duplex
asynchronous communication, which was much more desirable than 1200
bps half-duplex operation for timesharing terminal usage (but not,
perhaps, for IMP-to-IMP connections!).  See, for example, the Stanford
Artificial Intelligence Laboratory (SAIL) history Web site:

http://www-db.stanford.edu/pub/voy/museum/pictures/display/3-2-Modem.htm

	1200 bps full-duplex modems for ordinary phone lines were
commercially available at a "reasonable" price (under $1000) in 1976,
according to the above-cited SAIL Web site.

	I used 1200/150 full-duplex asynchronous modems to dialin to
UCLA's Center for Computer-based Behavioral Studies (UCLA/CCBS, an
ARPAnet node) from home in 1974 (or earlier?).  I used both a TI
Silent 700 terminal and one of CCBS's special storage tube display
terminals.  Switching from a 300 bps full-duplex modem to a 1200/150
full-duplex modem, I immediately noticed the reduced uplink bandwidth
when using the TI Silent 700: my Silent 700 didn't buffer keystrokes,
and I could type alternate-hand character pairs faster than they could
be transmitted, causing the second keystroke to be lost.  I had to
slow my typing rate to adapt to the "higher speed" modem.  I don't
think I had the same problem with the storage tube terminal; it
buffered keystrokes, I believe.  I can almost recall the schematic.
Somewhere in storage, I have a manual...  :-)

<end tangential reminiscence>

					Craig Milo Rogers




More information about the Internet-history mailing list