Birth of the IBM 7070 (original) (raw)

People sometimes assume that large corporations (and surely IBM!) make careful decisions. Not always.

IBM, around 1958 or 1959, wanted a new computer system for their "commercial" customers. The existing systems for this area were the IBM 705 and the 650, the latter retrofitted with magnetic tape that such customers used heavily.

IBM wanted a single type, but two had been designed. At Endicott, B. O. Evans (later chief of the IBM 360 project) had designed the 660 as a follow-on. At Poughkeepsie, Rex Rice had designed the 750 as the successor to the 705. As I had written software systems for both 650 and 705, they designated me to evaluate. I asked Tom Glans to write part of the 705PRINT Isystem for the 750. It checked out well.

But he confirmed a problem with the proposed 660. The tape units of the 705 were responsive to a character called a "Group Mark". As a character in its own right it could participate in sorting operations, and it was used much in program logic. The 650 did not have a Group Mark in its character set. And it was planned to use customers' existing databases from both the 650 and the 705 with the new machine.

The character set for the 650 had a problem. The machine itself used a decimal bi-quinary code. That is, 5 lights visible on the console said whether it was a 0, 1, 2, 3, or 4. A 6th light said, if ON, to add a 5 to these. A pair of them gave 100 combinations, assigned and used as both characters and operation codes. Thus the collating or sorting process had a problem. The character and operation code set was full-up! No room to add a Group Mark!

I tried for three and a half months to convince Evans that a 2-digit code could not possibly accommodate a "19 1/2", which is where the 705 Group Mark had to fit in order to keep the same collating sequence as the 705. This problem eventually killed all compatibility of instruction repertoire and ordered data.

Thus it was that Applied Programming recommended the 750. Unfortunately the programmers of those days, not yet emancipated, were not allowed to sit in the highest councils of decision at IBM. Our decision had to be relayed through Roger Bury, Manager of Education. Sadly, Roger somehow got it backward (he probably forgot to take notes), and said that Applied Programming wanted the 660. Seems as though everyone said that_that_ was a good basis for decision, even if the programmers were not allowed in the meeting.

And that's how the IBM 7070 came into being!

It's not really laughable. The 7070 begat the 7074, and it is rumored that one of these is still chunking away in some IRS office, even though it is a 6-bit machine, utterly incompatible with ASCII or even EBCDIC.