How the Bell System Missed the Internet 6
The Bell System ACS system was the largest privately financed R&D project in the history of mankind as of 1980. It was designed to have been the network that connected every home and business with data. The Bell System voice network was being crushed under a heavy load of dataphone calls. These calls were bringing analog switches to their knees. Companies would dial up a connection and leave it connected for 6 or 8 hours. In cities with flat-rate calling companies would dial-up a connection and leave it connected for days on end. The phone network was not designed for that. The Bell System needed a data switching Central Office and ACS was the answer.
Championed by Billy Oliver, the Steve Jobs” of the Bell System, a parallel data network was going to be built to off-load data from the voice POTS network. People would dial a 10-digit DTN (Digital Telephone Number) and be connected via packet switching. They would pay by the packet, rather than by the minute. The great thing was that calling between the networks would be transparent. An ACS customer could dial up a POTS customer and the ACS system would route the packets to the far end, and then automatically use a modem pool to reach the analog POTS customer.
The problem was that the ACS system just didn’t work. It was slow and unreliable. And, I had delivered an unambiguous memo to AT&T Long Lines that the project was a catastrophe. I believe I even mentioned that ACS could never be made to work because of many defects in the design. That memo made it up to Billy Oliver’s desk and an executive demonstration of ACS was scheduled.
The demonstration was a smashing success, contrary to my memo.
Following the successful executive demonstration I went back to Bell labs, as usual. It was time for some more system testing on my part. What happened that suddenly allows the ACS system to start working when just before it couldn’t do anything?
I went to the usual test terminals and initiated a connection. Presto! A swift connection when just a few days earlier it would take 30-seconds when the system was idle. Typing on one terminal would appear on the other almost in real-time. Yes, indeed, the system did appear to be working well. As Lilly Tomlin would say: “A gracious good afternoon to you!”
Then, as I stared through the glass and into the adjoining central office data center I noticed the line card bay. The LED indicators showed that it wasn’t even initialized. I walked over and powered it down.
The immediate thoughts are “What error have I made?” Perhaps all along I was looking at the wrong line cards, or perhaps the wrong lights. You know the feeling when you are working on something and then realize that your earlier observations may have been wrong. My mind was racing.
I continued to power parts of the ACS Central Office down. One by one, frames of equipment was shut down and yet the ACS system continued to work. Even with the main DEC PDP 11/34 and 11/70 systems powered down my connection remained responsive.
Even powered down, the system worked perfectly! I have never yet seen a computer system that performs flawlessly when powered down.
So I decided to follow the cables from the terminals to see what was going on. The terminal cables dropped into the raised floor of the data center (remember when we used to have raised floors, before fire codes basically banned them?) And, then the wires joined a large bundle and then they went to the floor downstairs. Downstairs there was a DEC PDP 11/70 sitting by its lonesome. And, the terminal wire simply went into a normal serial port on a port multiplexer. The same was true with all of the terminals upstairs. None of the terminals were actually wired to the ACS system. They came to this secret PDP 11/70 on the floor below.
Running on this mysterious PDP 11/70 system was a program I had not seen before. After some research I was to find that this program had been cobbled together in the prior week to emulate the ACS system. It simulated the command structure as ACS and it simulated calls being established. It even transmitted characters received from one port to the one that was called.
Of course, this ACS simulator didn’t use any of the ACS hardware. And, it could only switch calls between the 16 ports connected to the 11/70. And, it really didn’t accomplish anything. It was quite simply…a fraud. Bell Laboratories had completely faked the ACS demonstration to the senior executives of AT&T Long Lines!
Yes, the demonstration of ACS to the AT&T executives was a complete fake. And, now I knew how.