I think the gp was confused with other devices. Colossus was indeed a computer by most definitions. I think the poster winced it up with the Bombe or other systems - not surprising because colossus wasn’t really known for many years. (It was secret into the 1970s iirc)

Other devices would calculate but not store instructions. The common ones you see are the fire directors on naval ships, which were analog “computers”, but single purpose.

▲

Animats 7 days ago | parent [-]

By "computer" I mean what we call a computer today - a stored program general purpose electronic digital computer.

There were many early machines which checked some, but not all, of those boxes. IBM's electronic multiplier. The Harvard Mark I. The SSEC. Colossus. Reservisor. Western Electric Plan 55-A. General Railway Signal's NX. The Bell Labs Complex Calculator. The Automatic Odds option for racetrack totalizators. The Mathatron. All of those machines did something that resembled computation.

The late 1940s, 1950s, and 1960s were full of strange special-purpose electronic digital hardware that didn't quite make it to a computer, because the parts count to get to a general purpose machine was too high.

Then came microprocessors, and it became cheaper to use general purpose microprocessors in dedicated applications. Now all those weird machines are forgotten.

Here's a brochure from Teleregister, which built custom special purpose systems for railroads, the military, airlines, stock exchanges, and such, from before WWII into the 1960s. There's no computer in those things, but a lot of electronics.

	▲	adrian_b 7 days ago \| parent [-]
		The Harvard Mark I and its successors, and especially the IBM SSEC were "stored program general purpose computers". Mark I was an electro-mechanical computer, while SSEC was hybrid, including both electro-mechanical parts and parts with vacuum tubes. For a few years, IBM's SSEC was the world's most powerful "supercomputer", and it has solved a great number of diverse problems. SSEC had some advanced features that have been introduced in fully electronic computers only about a decade later, e.g. pipelined instruction execution (to compensate for its slow circuits). What you mean is that none of your examples was a von Neumann computer, i.e. where there is a common memory for program storage and data storage, enabling the computer to create or modify programs by itself. Obviously the common memory was an essential element for the evolution of electronic computers, enabling many features that were impossible when the programs were stored separately, on a ROM such as punched tape. However, saying just "stored program" also covers the case when the program is stored in a separate ROM, as it may still be the case for a microcontroller, though nowadays most of them store the program in an alterable flash memory.