First we need a little background. Most CD-i players have two input ports, usually one on the back and one on the front. Newer players support the use of a splitter cable to connect two peripherals to the front port, in reality these players have two "front" ports that are smartly wired together in a single connector (dual front port support is a feature of the IKAT coprocessor in the newer CD-i players, specifically Mono-III and later).
However, front and back ports are not functionally identical. Back ports are fully bidirectional RS232 ports supporting both serial input and serial output; they are invariably connected directly to the on-chip UART port of the 68070 processor. Front ports are unidirectional; they can only receive serial input and are connected via the SLAVE or IKAT coprocessor (an 68HC05).
The SLAVE processor fully supports the "K" protocol on its single input port, but the IKAT processor only supports it on its secondary port (the one that only becomes accessible when a splitter cable is used). There is no good technical reason for this limitation as far as I know: both ports are supported by software in the IKAT and it should have been easy to support the "K" protocol on both.
Hence the following requirement from the CD-i KeyControl manual:
Splitter cable is required for models CDI 210/40, CDI210/60 and CDI 470/00 and up. (you can see the full manual here).
There is however a complication with the CD-i 450 and FW 380i players (and their variations the CD-i 550 and Magnavox CDI 5000 and CDI 9000): they do not have a back port! Instead one of the two "ports" that make up the front port is connected to the 68070 and thus bidirectional, the other is unidirectional. For some reason the secondary port was chosen to be the bidirectional one, possibly to lessen the load in the main 68070 processor in the usual case or for fail-safety reasons (if a user blows up a port it would tend to be the primary one, which would leave the 68070 port accessible for diagnostics).
The CD-i modem has to be connected to a bidirectional input port, otherwise the CD-i player would not be able to send out data. This means the back port on those players that have it, or the secondary front port on those that do not.
From the above you can deduce that this leaves players without a back port without the possibility to connect a K-mode keyboard while also using a modem! That means that the CD-i 450 and FW 380i players (and their variations) could not be used in the modem + keyboard configuration, which was unacceptable for both marketing and philosophical reasons.
So a workaround was devised. By having the CD-i KeyControl masquerade as a graphical tablet, it could be supported on the primary input port of the IKAT. The coprocessor dutifully reports the "tablet" data to the main processor, which then (using a custom driver that was included on the application disc) transforms it back into keyboard data that the CD-i application can process in the normal way.
When both a modem and a CD-i KeyControl are connected to a CD-i 450 or the FW 380i player, there is no possibility for connecting a wired pointing device (both input ports are in use). These players have no remote (infrared) capability which would make it impossible to "point and click". To remedy this, another workaround was devised. In the T-mode, the function keys F3 to F6 act as cursor keys while the F1 and F2 keys act as button and button 2, respectively.
[Thanks, cdifan]