🚀 go-pugleaf

In article <11347@life.ai.mit.edu> melnik@gnu.ai.mit.edu (Ofer Melnik) writes:
+---------------
| A group of friends and I are in the process of creating an intelligent
| data acquisition system...  It consists of one master (an MSDOS machine)
| and up to 256 slave terminals...  The flow of communications is basiclly
| polling; The master repeatedly polls each slave,...
| As we see it asynchronous communications is the only viable form...
| The first being obviously that we have no way to send a clock over the line.
|  
| What protocols would you recomend for our applications?? Any infrmation
| would be greatly appreciated.
+---------------

I would recommend using DDCMP (the DEC link-level protocol) in polled
multi-drop mode, with multi-drop modems. This is what we did some 15+
years ago at Digital Communcations Associates to interconnect our stat
muxes when we had to use multi-drop lines. (Some of our customers were
replacing existing multi-drop IBM terminals with our gear -- we could
change the terminals but couldn't change the line or the modems.)

DDCMP is fairly easy to implement, and can be done in a small amount of
memory. (We did it in an 8K PDP-8, *including* the async drivers and
stat-mux functions for multiple terminals/node.) It explicitly supports
multi-drop, including letting the master bury an <ACK> to one slave in
a data packet to another (or something like that).

If the systems are reasonably close, you can simulate multi-drop modems
with a diode/resistor hack between the transmit line of each slave and
the common line back to the master. (This is what we did in-house when
testing.)


-Rob

-----
Rob Warnock, MS-9U/510		rpw3@sgi.com		rpw3@pei.com
Silicon Graphics, Inc.		(415)335-1673		Protocol Engines, Inc.
2011 N. Shoreline Blvd.
Mountain View, CA  94039-7311