Ok - so I'm going nowhere - thought that would be the case.
So lets open it out to a 'robot bus' - for example using single/dual inline headers/sockets to 'stack' boards on top of each other (by clipping together), or alongside each other (with a cable). Assuming the pins were organised: Gnd, 5v, Motor Supply, Uart#1, I2C, IO Port A, B, C, D, E etc so that the least common pins (ie bigger controllers) were on the outside then a small bot could have a smaller bus. A 'brain' board could then be changed between a smaller ATMega8 and a 40 pin ATMega32 with no other changes.
Why am I hung up about this - making electronic stuff modular and compatible with a larger number of processors.
This may also be impossible. So the only standard is 'screw size'?
What about headers/sockets? A photosensor, or servo may just need the normal 3 pins (gnd, +ve, signal) from the $50 header but what about a board that needs say: gnd, +ve, and 3 I/O pins. You could use 3 servo leads for the IO pins but each one also does gnd and +ve so there is a lot of duplication and excess wire. Or do you have a power supply (gnd, +ve) cable and then a 3 pin header that goes to 3 IO pin signals. The std Molex headers aren't polarised so can let you connected things back to front. So should there be a standard for power, vs UART, vs IO: ie Molex, or KK, or shrouded headers. My main reason for suggesting something is that when I go back to a board I designed 6 months eariler then I have to resort to the schematic to work out how to connect it even though I designed the thing!!!