brian@sdcsvax.UCSD.EDU (Brian Kantor) (01/12/88)
I actually built one of these from a 3B2 - we just took one of our spare 3B2s and loaded it up with serial ports - it will hold 16, or 12 if you want the Ethernet card in it too. No logins were started on the ports - they were turned 'off' in the /etc/inittab file. Each serial port was read in turn by a program that had opened them with O_NDELAY so that reads wouldn't block. Each port that was sampled had its own buffer, and characters were accumulated in the buffer until either a CR, LF, or FF was encountered, or until there had been no further characters for (I think) 10 seconds or so. Then the buffer was prefixed with the machine name and timestamp, then sent to the hardcopy printer attached to the system and logged to disk. It worked fine using a couple of terminals to test it. I never got around to splicing it into the various consoles in our machine room to test it under actual working conditions. If it had worked, I was going to replace all the individual consoles with a single console and a switch. That machine is gone now, and I didn't keep the program, but it's relatively trivial to write, if you care to duplicate the experiment. A few years ago, I'd designed a similar system based on an S-100 Z80 system with a couple of 8-port serial boards, but I couldn't convince the management types around there to spring for the $800 or so it would have cost to build it. Brian Kantor UCSD Office of Academic Computing Academic Network Operations Group UCSD B-028, La Jolla, CA 92093 USA