It's Mode 0 that I'm trying to fix. If it worked properly, one could use that mode to sample data at nearly any rate, limited by Nyquist, of course, despite the rate of the MCU. Further, one could communicate with devices that are pretty difficult to handle any more, e.g. floppy disks, tape drives, etc.

For now, the only way to do those things is by bit-banging, which is rate limited by the MCU's execution rate.

The serial port potentially offers more flexibility, less the limitations.

Remember the initial question. How badly would one miss having to clear the RI/TI bits if they were automatically cleared when you read or write, respectively, the serial port's data buffer. That's what you do immediately afterward anyway, isn't it?

RE