Mixing telephones and computers

One of the communication facilities we were allocated for the RISOS project was a number of auto dialed telephones. We had at least four and perhaps five of these telephones, one of which was on an outward going WATTs (free long distance) line. We had programmed our PDP-11/45 computer (with its "RATS" operating system) to dial these telephones and make connections to whatever computer answered. There is a story of how I used that WATTs line to stay connected to e-mail while traveling during a crisis ("Road Warrior") on this service award page. This page tells a somewhat related horror story of mixing computers with telephones.

Most of the time the RISOS dialers were used for calling ARPA network "TIPs" (see IMP) such as the one at NASA Ames. We did have telephone numbers on our RISOS RATS system also for some other computers (some WWMCCS and EXEC8 systems, and some dial up access to some IBM systems running VS2 and VM) systems that we needed to call from RISOS. We had written software that would support a list of telephone numbers for any given access we required. For example, if we wanted to call an ARPAnet TIP it would try multiple try multiple telephone numbers that we had programmed into a list.

This mechanism seemed to work pretty well. However, it did require some maintenance. If, for example, some telephone number was retired we didn't want our system to keep trying it, always wasting time trying and failing on a retired telephone number. At some point I'd been noticing that for about 6-8 weeks when our system tried to call the NASA Ames TIP it was always failing on the first number it tried. I thought this might be one of those situations were a number had been retired so I decided to look into it. The first thing I did was to try calling the number manually. In this case, to my surprise, the call went through to that first number and I got a modem answer. I couldn't understand this because when the computer called the number it always failed to get through. I tested again with the computer - failure. I tried manually and the call went through. I must have puzzled over this for a day or two. The auto dialers that we had in those days were "pulse" dialers. They functioned by breaking and remaking the voltage on the telephone line rapidly they way the old manual telephones worked. It happened that on our auto dialed telephones there was a light that would blink as the circuit was disconnected and reconnected. In effect I could "watch" as the telephone number was dialed. I watched carefully as the number for the Ames TIP was called. After a few tries I notied that one of the digits that was supposed to be a 5 (one, zero, one in binary) was instead being dialed as 4 (one zero zero). I spent a little time researching the auto dialers and determined that they had individual registers for each digit. These registers were loaded with the digits from the telephone number and then the dial would sequence through the digits of the telephone number by reading out each digit and breaking and remaking the circuit appropriately.

Gradually it dawned on me (after more failed auto dials to the Ames TIP) that there must be a bad bit in the register for that one digit. This would only cause a problem with this one dialer and even then only if the number being dialed in that partcular digit was odd. Since the first number on the list for the Ames TIP had a "5" in that digit the dialer would always fail when dialing that number.

At this point it occurred to me to call the number that the dialer had been incorrectly dialing. When I called it a woman's voice answered. I spoke to her briefly. She was very angry! She had been complaining to the telephone company for the last month about all the spurious calls (nobody at the other end as one expects with a modem call) she had been getting. Our computer must have been calling the Ames TIP on average 3-5 times a day. From our viewpoint it was a bit of a bother that it always took a little longer because that first number failed to get through, but after the one failure the auto dialer mechanism would try the next number and our call would go through. From the viewpoint of this woman and her family, however, (some home in the Mountain View area) she was getting spurious calls that averaged many per day at all hours that would have nobody at the other end. It was no wonder that she was upset!

I immediately took that telephone number out of our list and called the repair folks from AT&T to fix the dialer. They commented that it was quite unusual to get a repair call where the customer could tell them so exactly what was wrong with a dialer (one bit dropped in one digit). I'm sure that woman was relieved when her phantom calls stopped. From my perspective I felt I'd learned a lesson about the danger of mixing computers with telephones.

--Jed