1 every time u add a component to a system you double the complexity. If you build automated regression testing it becomes easy to see the impact as test runtime grows exponentially. So the definition of the interface between components must be able to isolate them from one another. We saw the consequences in PC land with cards and dip switches and irq settings and plugnplay and even pcpartsbuilder. Vs say USB c.
Like there is a spectrum between pluggable sidecars and Bluetooth. In between is open isa , scsi, adb, usb, ethernet, wireless. Each one has tradeoffs in cost per module , cost of manufacture, cost of testing, physical form (space) and user experience. The market decided where the tradeoffs mattered most and it basically rejected every single side car variation ever produced. Its a niche thing that always remains niche due to suboptimal tradeoffs for most use cases
And i bet it is mainly a version of the premature optimization issue. A side car is spending enormous manufacturing resources to optimize something people will do extremely rarely… plug parts of a cpu box machine together. If you need to frequently plug and unplug…. You are better off with some interface that uses cables and … like sequential io vs random io you wouldnt want to have to get the module from the middle of your stack. If you dont need to plug very often…. Then you wind up paying for some enormous amount of plastic molding work and pcb design and testing for something you do maybe once a year or less. Better off to just buy bare boards like in PC land. But even if you have fancy plastic cases on modules that doesnt solve the software side. So ironically the thing that makes usb great isnt just the physical design but the software isolation between subsystems .
A sidecar type design is literally just exposing the bus, and yes, it rejected exposing the bus externally over "just" providing a variety of cases to take cards internally. On the Amiga, it was literally exposing the raw CPU bus.
Most of the 1980's sidecar designs existed to provide options at a point where they were competing at price points and in market segments where that was not viable. At the same time the price points for peripherals made the extra plastic a rounding error at that point in time. I think most of these designs failed largely because the moment you got people thinking about how many extra things they might assemble to get what they want, they'd look for a machine that provides more of that out of the box instead.
Having a sidecar felt like an "escape hatch" so you didn't feel locked in when buying a fairly basic machine. If you think there's even a possibility you might need 3+, you'd be looking at a big box machine from the start.
then the question becomes.... why was the original machine not simply expandable with extra sidecars, why couldn't they upgrade their TI 99 4/a or PC Jr with sidecars so it could do the same as a machine that provided "more out of the box".
is it because CPU freq kept doubling every 18 months or whatever? but i would argue, when that doubling stopped and we reached the 5 ghz limit, we didn't see sidecars make a comeback in computer land.
the closest thing we have is external GPU which is still, after all, cabled, not sidecard.
i am making a hypothesis again that the cost of upgrading a machine with sidecars is always going to be higher than just buying a machine that allows lower cost upgrades via slots or cables.
or in Apple worldview, its better 'overall' from user experience to just buy the nextgen which has very little or no expandability. and just guess what people actually need/want based on careful research and connection with customer base.
now i dont have good data on cost of sidecard vs other interconnects but it seems inherently higher to me. thanks for the response.
It's very simple: Because these machines were for the most part cost-cut to the very bone for people to be able to afford buying them at all.
When I bought my first Amiga, an Amiga 500 was the most I could afford. When I could afford to add a (sidecar) harddrive, that was all I could afford. If I were to wait until I could afford an Amiga 2000 and an internal harddrive, that might have been cheaper than the combination but it would have meant waiting a couple of years to get one at all.
These were the market realities these machines were built for. And, yes, that made it a lot less attractive for machines targeting business customers.
But I'll note not so much because of the cost of the boxes - my first harddrive cost almost as much as Amiga had cost. Used. It'd have been almost as expensive as an expansion card for a big box, because the controller was pretty much a full computer on it's own, and the drive itself was ridiculously expensive.
Like there is a spectrum between pluggable sidecars and Bluetooth. In between is open isa , scsi, adb, usb, ethernet, wireless. Each one has tradeoffs in cost per module , cost of manufacture, cost of testing, physical form (space) and user experience. The market decided where the tradeoffs mattered most and it basically rejected every single side car variation ever produced. Its a niche thing that always remains niche due to suboptimal tradeoffs for most use cases
And i bet it is mainly a version of the premature optimization issue. A side car is spending enormous manufacturing resources to optimize something people will do extremely rarely… plug parts of a cpu box machine together. If you need to frequently plug and unplug…. You are better off with some interface that uses cables and … like sequential io vs random io you wouldnt want to have to get the module from the middle of your stack. If you dont need to plug very often…. Then you wind up paying for some enormous amount of plastic molding work and pcb design and testing for something you do maybe once a year or less. Better off to just buy bare boards like in PC land. But even if you have fancy plastic cases on modules that doesnt solve the software side. So ironically the thing that makes usb great isnt just the physical design but the software isolation between subsystems .