>Yuo might also want to take a look at Section 6.1, "TIMER0 (TMR0)
>INTERRUPT", which says, in so many words, "The TMR0 module interrupt
>cannot wake the processor from SLEEP...."

To expand on that, the RTCC doesn't work at all during sleep, as the
synchronizer is driven by the internal PIC clock.  The prescaler (if
enabled) probably does continue to count, but the main register won't.

>If your RB0 pin is available, I'd tie it to the RA4/RTCC pin and
>configure the RB0 interrupt to occur when it sees the start bit.
>
There's really no need to use two pins, you could take all input through a
port B pin, and use the RTCC with internal clock to time the bits and cause
interrupts at the correct sample times.  Or, if the PIC has nothing to do
until it gets some more serial data, an ordinary software-timed receiver may
be all that's required.  At high baud rates, the (uncertain) time required
to re-start the oscillator may make it difficult to know where the center of
the start bit (that woke up the PIC) was and sample at the right time.
Sending a 'don't care' byte to wake up the PIC, then immediately following
it with real data may be a way around that.

-Mike