I'm not sure why doing these is so amusing, but I was compelled to do a few more.
First I did a left-handed / mirrored C64 case. This is not for anything in particular – I was just goofing around. As a leftie using a traditional 104 key keyboard feels a bit like swinging a baseball bat with the off hand, hence the flip.
I have honestly forgotten what the special C64 keys do and had to look it up. It's up to the editor in memory to interpret the keycodes, except for Restore which is not part of the keyboard but goes separately to an interrupt line – very useful for resetting bad code in a more reliable hardware kind of way.
I've added a pair of Alt keys. I think the C= key is already basically Alt, but for the PETSCII symbol layer, which I omitted in favour of implementing one of my compact keyboard layouts. Technically the C= key should strictly be an OS shortcut key I think. The C64 had dedicated @ and £ keys... kind of strange? Since BASIC uses e.g. A$="" for strings you'd think a primary $ key would make more sense.
In Sweden the less used [,£,] were turned into ÄÖÅ.
The leftie numpad is mirrored. One might think this causes confusion, but as a leftie I don't actually have any numpad muscle memory since I can't really use it, so a novel layout is alright. It makes more sense to put the very common 1 and 0 under the index finger. Left handed keyboards – the few that exist – sometimes mirror the entire numpad block. Arrow and page keys also go on the left.
----
Additionally, I did another Amiga variant. This one is based on the Ranger – an enigmatic, nebulous forerunner of AGA and perhaps what the Amiga lineup needed in the late '80s. The Ranger concept supposedly instead turned into ECS (or was ECS all along) which in turn turned into AGA. ECS is short for Enhanced Chip Set – an improvement over the Original Chip Set, but not many home users cared about the monitor-only resolutions or ultra-thin pixels. AGA was not up to snuff by the time it came out, and I think retro-restriction-wise it's less interesting to work with AGA than the more distinct OCS nowadays.
The Amiga 1000 and 500 used a regular Motorola 68000, which while internally a 32-bit chip, only exposed a 24-bit address bus. This still allowed it to address a whopping 16 MB though – far more than home users ever had. I had 6 MB in 93-94 and that was exceptional.
The 68020 used in e.g. the Amiga 1200 (edit: only Blizzard accelerated machines) made it possible to use up to 4 GB of memory, but this never happened in 1992 of course. The earlier Ranger project (mid-late '80s) however was probably utilizing the older 68010 which was more similar to the original 68000, except that it supported virtualization, vector offsets, allowing for some sandboxing and crash resilience. It was noticeably faster in certain kind of loops too. Anyways, I think it fits OCS era restrictions better, making it a good choice for an Amiga 500 era fantasy computer.
The Ranger design is rumoured to have used 2 MB of faster VRAM (dual port?) and some C= engineers were eying the 881 MMU. I don't really know how it would've worked exactly, but it would have enabled flickerless high resolutions and/or more colour depth.
Remnants of the high resolution modes can be seen in ECS and AGA where you can use strange RBG222 four colour modes. Here the RBG sliders are 2-bit each with 4 steps rather than 16, i.e. you can only use a palette of 64 pretty garish colours. And 4 colours total. This is actually sufficient for productivity software like CAD, Spreadsheet and and text editing. Some of the modes required a specific monitor type. The reason for the colour limitation was memory bandwidth. The more pixels a frame has, the more compact they have to be in memory in order to load in time. ECS has some greyscale modes and the saving there is probably not having to access the colour table – index is instead outputted directly as brightness. I suppose having a dedicated Palette DAC would be nice, but then it might be more difficult to manipulate the palette for neat flashing, cycling and copper effects.
Regardless, as things are, having both Chip and Fast RAM does speed things up. Chip RAM is used by the graphics and sound, whilst the CPU apparently benefits from Fast RAM. It might be possible to squeeze out some extra pixels by just using a higher clock and faster memory, with some changes in Agnus.
Personally I was never fond of the tall-pixel resolutions so I wouldn't support those. 400x300 might just be my favourite resolution (at 640x480 you kinda lose the pixels but it'd be nice for PC-98 ports). More sprites would have been nice too. More bitplanes (e.g. a 128 colour mode) isn't as interesting to me however – I think above 32 colours there are diminishing returns unless you're doing games with light-level effects.
My joypad here was based on the CD-TV one. The Amiga generally used one-button joysticks. The port kind of supported two or three buttons but in practice no one had any joysticks like that and many games were hardcoded for one button controls (plus keyboard). For my own keyboard a while back I just hooked the joypad into the keyboard matrix. A 3x4 matrix has to be scanned several times which makes polling slower (though free if the keyboard is already being polled), but one also gets 3x4=12 buttons out of 7 wires. No need to put shift registers in the pad or decoding logic in the computer. The disadvantage is needing to diode the entire keyboard, but that's not really a minus as it saves key ghosting headaches.
Edit: Added an RGB222 palette for clarification. The palette is not completely unusable and a little easier to understand if laid out as ramps.