Would it not have been cheaper for the video interface to be integrated into the host, obviating the need for memory and processing the peripheral?

@epl Erm - I did mention this, didn't I? The memory would be needed anyway, after all, the picture had to be stored for display. There is nothing to be saved, but within the computer it would take away much needed RAM from CPU usage, wouldn't it? Address space is an extreme scare resource. This wasn't 64 bit address space like today.

It's hard to see how two devices is more economical than one even if the assumption would be robust (and it seems not so to me) that the total memory required for both configurations is the same. At minimum, achieving the same functionality without an additional processor, as appearing in later terminals, would seem to be a big savings.

@epl You're describing the Apple I and asking why it was invented when it was and not sooner.

@snips-n-snails: I suppose the question might be framed as you have done, but I would rather see the video adapter as a choice, not an invention, as plainly Job's team was not the first to conceive of it, or even to build it.

@epl You may want to see this from the computer manufacturers point of view. If you would start to write a new program, will you write anything from scratch, or use existing components? I guess using existing components like it is - and the same is by first designing just the computer. Even if from a users perspective an all in one computer might be more useful.

@Raffzahn The evolution of computer equipment had already become a fiercely competitive an innovative space, with the market demanding new capabilities at lower cost every several years. I doubt complacency was the reason for such developments not occurring until decades after they would have been feasible.

@epl If you talk about decades of development, then please specify the exact time range you have in mind - and also tell where there is a fierce competition. First (somewhat general usable) micro computers came in 1975 - and by 1977/78 all in one units like Apple II, TRS-80 or PET were available. 2-3 years are not decades. And before that, computers where most of the time MULTI user machines. Multi-user need by default terminals - not at least because the allow longer in house connections than a few meter.

Raffzanh: Is it not clear from context that the relevant period would be bounded roughly by video terminals becoming prevalent, and desktop systems, which have integrated video adapters, replacing microcomputers and minicomputers?

@epl No. (Today's) Desktop computers ARE microcomputers. There were a few non micro desktop computers (like mentioned) in the early 1970s. Except for them EVERYTHING was mainframe or mini. And all of that (except for very special use cases) was terminal based and serving multiple users at a time. How else to do this than with terminals? The one computer <-> one user relation is a thing that only came after micro computers became available after 1975. And micros with build in video became available short after (ca. 1977).

I have no wish to debate some canonical understanding of terminology. I am following a practical understanding that the desktop computer is the name given to the successor of the microcomputer, even if no hard boundary is identified. Apples and PCs were desktops, not microcomputers, in the popular lexicon, which is completely adequate for the current discussion.

@epl Erm, sorry, PCs are by definition micro computers. It would be helpful to stay with established terms and definitions. Otherwise it might be better to close that question as it seems to be self serving and denying an serious answer.

As I indicated, I am content that the vernacular terminology is expressive enough to convey the essence of the question. I have no wish to quibble over definitions.

@epl That people do not go well with your wording should be telling, shouldn't it? Be assured, there is quite some knowledge about this topic on this site, but private, made up and non standard definition, paired with unwillingness to clarify might not be the best way to conceive a meaningful answer.

I would think that using 1960s technology, the hardware to connect a computer to a ten or so CRTs each showing a separate 80x24 text screen would be far less than twelve times the hardware required for one such display terminal. Use one main display memory and then have a bunch of boards, each with seven acoustic delay lines (to delay one text line worth of data by one scan line) and a diode ROM for one line of the font, and an output transistor for each terminal to route the output.

@supercat Well, that's in part what IBM did with the 2260 - one controller containing most logic and a set of rather less effort viewing stations within a few meters distance. Still, a one off example, as it only pays with large installations. More important, terminals weren't a thing in the 1960s. they only came to notable use after 1970.

"all the processors stuck in anything from Furbys to CNC machinery"... Now I'm thinking about CNC machines being operated by Furbys. :)

@another-dave no, it was purpose build, using standard components Of course, a terminal controller does need some processing capabilities, as he needs to interpret mainframe output, handle keyboard input and so on, still it's way apart from a mini.

@Raffzahn: I wonder if any such systems ever interfaced to display terminals using RF-modulated video. While RF-modulated video is inferior to baseband video at short distances, but many channels can easily be combined on a single cable and sent for miles. If one wanted to have a computer shared by people in a couple of buildings, sending video between buildings using an RF coax and multiplexing keyboards would seem like it could be easier than trying to convey display content as data.

@supercat Well, we did exactly that in 1976/77 to build cheap ass terminals for an inhouse solution based on a scrap metal /360. We took a disk controller (main frame, that's a whole rack) removed everything but the mainframe interface and added a logic to store 'disk' blocks in video buffers. 16 at first. Each buffer made 16 lines of 32 characters, 6 bit each. Output was modulated onto 16 different carriers (channels) and distributed thruout the department. Input was build from telephone keypads using one of the first single chip touch tone solutions - cranked up to 500 kHz to ... 1/+

@supercat ... allow fast typing. Displays were lowest price B&W tellies from a department store. To save further cost, two keyboards shared the same output channel. We build the whole system for less (not counting time) than a single terminal controller plus two terminals would have cost. But yeah, very limited, very special, but the users loved it. In fact, when we were able to put a terminal onto every desk in the mid-1980s, many users refused, or wanted to keep their 'Gucki' in parallel. We had to support the odd system way into the 1990s. 2/-

@Raffzahn: That sounds neat. How much circuitry was shared among the video outputs?

@supercat The interface? After all, each had to hold their own the buffer (512x6 bit), character ROM, counters, video logic, modulator. So not much to have in common there. The disk interface handled all mainframe interaction, offering buffering a block from it and handing it to the destination buffer using the output counters of each. Keyboard input was turned into disk status messages. No micro processor, all discrete random logic.

@Raffzahn: I would think that if one were generating a dozen displays, one could stagger the video generation so that each row of the character generator ROM would only be used by one display at a time, only one line would need to be fetched from the main video buffer every scan line, and all other temporary storage would be handled using delay lines. A big advantage in an era when character ROM would be a diode matrix, but not so big when it's a single chip.

@supercat One can doe quite a lot. But belie me, that design was made by two very good (the best I ever met), old fashioned engineers, sparing every gate possible. your idea would have added much complexity in case of partitial failure as well as expansion. Also, character ROMS were among the very first single ship devices, so really cheap by mid 1970s - even Woz used one much like the one we did.

@Raffzahn: Yeah, I can imagine that by the 1970s the advantages of a heavily-multiplexed design would have been diminished severely, but I would think that in the 1960s they could have been significant. As for expansion, I'd expect that one would design each assembly to handle some number of monitors (e.g. 12) and if one needs more monitors one would add another assembly for each group of 12. In the Apollo mission control, some of the computers fed multiple displays, but I don't know how the hardware was implemented.