I have to explain why the driver didn't work with the EGI when I removed the :PB command on formerly versions.
I think it was because the mouse driver never passed the beta-testing phase. The driver needed to be loaded, then the mouse device was linked but not initialised. Then the user(from Basic or the Paintbox program) needed to type the :PB command to initialise it.
The coders of the driver wanted to initialise only the mouse, not the entire system, so the initialisation ended with registers "c"=0 and "a" loaded with the memory page where the driver was loaded. This is why the initialisation command produced an error, 9000+memory page. The characteristic was only needed for testing purposes.
This behaviour was not important with the Paintbox program, because the resulting codes where ignored, but when the EGI loads, it takes over the entire system and initialises it sending the "Action Code 8" to all devices.
My error was that I only changed the :PB command by an "Action Code 8" entry, not changing the results on registers "c" and "a" at the exit. The value on "a" can be 0 or an error code, but "c" must be 8 if we want the initialisation to be spread to all the devices(EXOS Kernel Specification 9.2.8 ). As a result, the drivers initialisation call always was stopped when it arrived to the mouse driver, so the EXDOS interface and other important devices ceased to be initialized.
It can be that the EGI coder, Rhodri Cusack, knew about that issue but didn't want to correct the Mouse driver by himself, as it wasn't his own creation.