Trying to get more into the specifics of the BBC connection, there is a substantial hint in the user guide:
However only 5 bits of the [user] port, and CB1, CB2 are used: This leaves bits 1,3 and 4 available for other uses.
Which is backed up by the schematic provided by Simon Inns in the doco for SmallyMouse2; comparing that to the user port's pinout the mouse provides inputs:
- X1 on the VIA's CB1;
- Y1 on the VIA's CB2;
- X2 on bit 0 of the VIA's port B;
- Y2 on bit 2 of the VIA's port B; and
- the three buttons, left, middle, right on bits 5–7 of the VIA's port B respectively.
So the value you found in page 0 was that of the input to port B: three bits corresponded to the buttons, and two of the others related to motion in X and Y as you suspected.
As per Justme's answer, mice of the era were quadrature encoded, producing two outputs per axis. On the x axis they are conventionally called X1 and X2. Both carry the output of a square wave that has a frequency proportional to the mouse velocity. They differ in that they're ninety degrees out of phase.
So during a fast movement then they might output a signal like:
X1 = 00110011
X2 = 01100110
Now imagine the user changed direction partway through, so that the pattern reversed itself:
X1 = 00110011 | 11001100
X2 = 01100110 | 01100110
If they had halved their speed when they reversed direction:
X1 = 00110011 | 11110000
X2 = 01100110 | 00111100
So a cheap way to track such a mouse is:
- Set up an interrupt for each time X1 transitions from 0 to 1.
- When X1 transitions, note a pixel's movement and sample X2.
- If X2 is 1, move to the right. Otherwise move to the left.
(And for greater precision, also interrupt on X1 transitions in the other direction, but for those reverse your X2 test, and also interrupt on X2 transitions, doing the direction test on X1)
The 6522 in the BBC is capable of raising an interrupt upon transitions in
CB2, so that is how software is alerted to mouse movements without constantly polling.
You can ask the 6522 to interrupt on either an upward transition or a downward one, but not both, so I would imagine the implementation is the simplest that I've described: transitions on X1 trigger a cursor movement, and sampling of the other signal picks between a left or right movement.
It'll likely be transitions in one direction only (e.g. low to high, ignoring high to low) as although you could reprogram the 6522 on every transition I'll wager that they didn't out of a preference for lower interrupt processing time, and because the BBC Micro screen is already low resolution enough that input can be relatively low precision.
So the only thing you didn't spot in your investigation was the interrupts.