Data communications over telegraph circuits 1899

Jim Haynes

Interesting point. I sensed my time-sharing systems were locally based in cities. In 1970 there was a small company "Community Computer Corp" that used an HP2000 to provide time sharing. In those days (60s and 70s) more businesses were in the cities than suburbs and could hook up with a local phone call. I don't think the time-sharing computers of that era could handle that many simultaneous active calls so they had to be locally anyway. Time sharing had a connect charge and that wasn't cheap. Most users prepared their work offline on paper tape, got online and did a run as fast as possible, then got offline to debug the results.

I think the TWX sale to WU was relatively late when toll rates were dropping. But I don't think WU actually got very much out of the deal since it still was paying AT&T to lease local loops.

I'm kind of surprised you'd say that. Depending on the time frame, my 1970s impression was (at least through my employer) that the Bell System was looking toward data communications as a huge growth industry for it, for both private line and switched. They introduced a wider variety of modern data phones, Teletype came out with the Dataspeed units, and so on. I think they expected a demand for private lines. Meanwhile, they were upgrading their own network to higher capacity carrier options like T1 and looking toward fibre and waveguide. Switches were going to ESS and than to digital ESS.

But, I wonder if they had only a few of the large companies with private wire networks. In comparison, Bell could be offering more pvt line services in a single small city than WU did nationwide. And that leads to my other puzzlement--if they did have those large companies as customers, how come they didn't build on that base with microwave?

Interesting point. I suspect they got burned. I don't know all the AUTODIN details, but I'm speculating here:

I'd say the physical life of a mainframe computer would be about 10 years, but technically obsolete in about 5 years. Given the very rapid technical changes of the 1960s, I suspect in about 5 years they found themselves with logically obsolete hardware, but conversion to something newer would've been extremely costly. I suspect in 10 years they had to convert to maintain reliability and that was really expensive.

IBM made two fateful decisions in the early 1960s: 1) It would provide hardware compatibility with older machines to newer machines so customers' programs could still run, indeed, run even better, without costly disruptive conversion. That support lasted for 35 years! 2) It would have a single unified architecture for its large and small and business and scientific machines. To this day IBM still sells a heck of a lot of computing horsepower and systems software that goes back to that 1962 design. Customers who bought IBM gear were generally protected from costly conversions.

I believe you mentioned earlier IBM's benefit from large economies of scale in manufacturing. I think that and IBM's unified architecture enabled to survive and even thrive while other computer makers have come and gone. When I worked with Univac gear, I realized how thin they were spread compared to IBM. Both mainframes need systems programmers to wrote compilers and operating systems, etc., but IBM gets a lot more 'wear' out of its people through economies of scale.

I suspect Western Union was doing a lot of unique research in computer technology searching to improve its business. However, things changed so fast that that research would be obsolete quickly. Further, WU just wasn't so big to spread the costs of research out of a large customer base. IBM figures out an improved circuit board and manufacturers many thousands of them. Western Union might make a few hundred.

(Similarly, when H&H was doing well, its central commissary and quality control worked well with many restaurants to support. But as the chain contracted it found itself with expensive wasted overhead).