Morton Goldberg wrote:
> On Jul 19, 2007, at 9:00 PM, John W. Kennedy wrote:
> 
>> Never had a 650 to work on, but read the manual once, a long, long, 
>> long time ago. My first programs were written for the 7070, which was 
>> the follow-on to the 650, but, since it used core, had a conventional 
>> instruction counter instead of a next-instruction-address field.
> 
> I wouldn't characterize any 70X/70X0 machine as a follow-on to the 650. 
> They were a mainframe series derived from the Whirlwind project, an IBM 
> military contract. The R&D tab for the 70Xs was picked-up by the US 
> taxpayer. The 70Xs were vacuum tube machines and were introduced at the 
> same time as the 650 -- the 70X0s were the later transistorized 
> versions.

A) There were a number of completely different 70x/70x0 lines, and, B) 
the Whirlwind was done by MIT; I cannot at this moment find any evidence 
that IBM was involved at all, although IBM later used parts of the 
abandoned Whirlwind II specs for the noncommercial AN/FSQ-7. Development 
of the 701 was certainly stimulated by the Korean War, but it and its 
successors were never intended specifically for government use.

> I don't remember a 707, but if there was a 7070, I would 
> expect a 707 existed as a predecessor. I did some Fortran programming on 
> a 709 and a 7040.

That's simply wrong. There was no 707, and the 7070 /was/ the follow-on 
to the 650, as any history of IBM's pre-360 computers will tell you. As 
with the 701->704, 702->705, and 1401->1410 transitions, the machines 
were not compatible, but general concepts and formats were. In this 
case, the 7070 retained the basic 650 word architecture of ten decimal 
digits plus a sign, and continued to use two-digit op-codes and 
four-digit addresses. But the four digits that the 650 used for the 
next-instruction address were replaced by two digits to select an index 
register and two digits to select a subfield of the data word.

> AFAIK, the 650 was developed as a business machine with IBM's own funds 
> by an entirely different engineering group than the Whirlwind group. 
> There is nothing in common between the two architectures. The 650 not 
> only used dual-address instructions, it wasn't even a binary machine. It 
> was a dead-end architecture.

No, the 7070, 7072, and 7074 were follow-ons to it.

The IBM pre-360 commercial lines (-> incompatible; => compatible):

The main scientific line
701->704=>709=>7090=>7094=>7094 II
                 |
                 .->7040=>7044
(The 7040 line was a cheaper subset of the 7090 line. It did /not/ 
derive directly from the 704, as some have incorrectly concluded from 
the number.)

The famous STRETCH supercomputer. Not truly a failure, but IBM lost 
money on it when they gave refunds because it wasn't as good as 
promised. Still, a lot of STRETCH concepts, reworked, went into the 360.
7030

Budget scientific and real-time.
1620=>1710
(The 1710 was a 1620 with command-and-control extras. There was also a 
one-off 1720.)

The main business line.
702->705=>705 II=>705 III=>7080

Cheaper business machines (and upward extensions).
1401=>1460
  | |
  | .->1440
  |
  .->1410=>7010
(The 1440 was an almost-compatible cheaper version of the 1401; the only 
incompatibility was in the handling of punched cards and printing. The 
1410 was an incompatible upward extension of the line; machine code was 
different, but carefully written assembler code could be portable. Two 
more systems in the line were the 1240 and the 1420, which were 
essentially 1401s with magnetic-ink reader/sorters in the same chassis.)

The first machine with a disk for business. Existing systems with added 
disks were more successful.
350

Budget general-purpose machines.
650->7070=>7072=>7074

Almost a PC. Too little and too late to get much of a market.
610


-- 
John W. Kennedy
If Bill Gates believes in "intelligent design", why can't he apply it to 
Windows?