I don't think that was Erik's point (although I don't know what it was indeed), and it's definitively not my point.

The reason why these chips attracted my attention is the integrated parallel host ("ISA"/"8042-style") interface, which I am missing in most of the microcontrollers (see the "kithchen-sink" thread a year ago). It's a simple one-byte FIFO with the appropriate signalling - nothing spectacular, but very useful if you need multiprocessor communication.

A simple example, there are a lots of ethernet gadgets around today, comprising from an mcu and an ethernet chip plus the needed passives, connectors etc. The mcu can perform all the low-level ethernet job plus a complete TCP/IP stack and provide an interface to a host, free of the unnecessary overhead (both in execution time, code and data memory) - e.g. for simply send and receive data. The host might be in a form of a SBC where parallel address/data bus is already present to connect memories and I/O devices, so why not to connect an another one in such a simple and straighforward way?

As I mentioned in the "kitchen sink" thread, I have already used similar asynchronous processor-to-processor interface, but it was built from discrete 74xxx's and had limited capability compared to what the "8042-style" interface provides (I had it one direction only, with no backward "buffer empty" signalling; the slave had a fixed time to pick up the "message").

Jan Waclawek