I envisage that the standard machine would offer just one monitor output, but that others could be added on as needed, e.g. on a bus. Some high-end software would require multiple monitors, but most software would just take advantage of it if available.
We already have the generic structure of the View in the Amiga. The View is the monitor, and has several ViewPorts. But since the View is a standard C structure, there is no reason why there should not be any number of them.
I guess that the main h/w problem will be memory contention - 8 monitors wanting memory access at some high rate causes quite a bottleneck. However, the Amiga already has the solution to this in its chip memory concept. Therefore: have one chip memory per monitor. (And extra flag in AllocMem to say which chipmem we want.) (Of course, 16Mb+ of chip mem per monitor, not 2Mb.)
Each chip memory would be mapped to a different address, and each would have its own Copper (or equivalent) plus its own video hardware. Each also to have their own set of sprites. (If we keep the architecture similar to what we have at present - but of course, with up-to-date electronics, we could have more-and-better versions of them.) It should be possible to design a suitable bus-like architecture that would detect how many are available, and have them all just running away independently.
(In Annotator, you draw rectangles etc. onto the image to mark things of interest, and note down information about them. e.g. an old wedding photograph: mark all the faces and say who they are, then draw the relationships between them. e.g. picture of Mars surface, mark significant rocks and other things, and record data about them e.g. their density and level of erosion.)
In any case, the concept is that the image itself is not marked, so never gets corrupted when you alter the markings. Instead, it acts like a transparent sheet on top of it. Made possible by dual playfield. But I want more.
So does computer aided design and geographic information systems. They have the idea of 'map layers', laid on top of one another.
Multiple playfields would make this great. To do all this on a single playfield (as on most platforms) means cumbersome redrawings when the layers are moved or altered.
But, please make it better than current Amiga's version, which is limited to only 4 planes per field and these have to be interleaved. So your fields must have approximately the same depth. No. I want to allow one with 256 colours and the next with only 2, then the next with 32, etc., all over a HAM8 background. (Or 24-bit, if you like; I am just going by standard Amiga gfx, but don't have any real reason to be restricted to it.)
Why do I want several each with 256 colours? Because I want one GIS-layer for contours, with gradual shading of browns to indicate heights, then another of roads, with gradual shading of reds and yellows to indicate traffic volumes, and another with shadings of greens to indicate areas of special conservation value, but of different values. Got the idea?
The software concepts are already there. The ViewPort can have two RasInfo structures, linked with a chain of pointers. No limit in the software.
The hardware should now be feasible. What is needed is that each playfield has its own raster-driving hardware and there is a priority chain among them, a bit like the SCSI chain. The hardware limit will be on the sorting out of these priorities fast enough, but current electronics should be able to allow up to 16 8-bit fields or so.
We could even make use of *current* Amiga chip technology, slow and limited as it is, to provide a rather super machine. Specialized, I agree, but there would be nothing else quite like it. A high added-value niche market.