|However, I have personally experienced problems with PICs (16F84)
|when attempting to use the internal protection diodes to limit i/o
|pin excursions. The processor operated in an undefined and
|intermittent manner. Changing the design so that it guaranteed the
|input pin would not sink/source protection diode current cured the
|problem totally. The current drawn was much smaller than the PIC
|design spec (ie about 1ma max). The application was a simple RS232
|input using a 10K ohm series resistor to the PIC pin.

I've used the input protection diodes on the PIC for low-current clamping
and have never had any problems, PROVIDED that I ensured that the input
did not pull VDD higher than spec.  Of course, RA4 has different rules,
but for most I/O's the clamps seem to work just fine.

I do wish, though, that the specs were a little more clearly written.  If
the specs are supposed to say what Microchip's ap-note suggests they mean,
it might be clearer if they were written:

DC characteristics:
  Input leakage: (?)uA

Operating conditions:
  Maximum input voltage: VDD + (?)V    (note 1)
  Minimum input voltage: Vss - (?)V    (note 1)
  Maximum input current: +/-(?)mA      (note 1)
  Maximum RA4 voltage:   (?)V          (note 2)
  Maximum RA4 current:   (?)mA         (note 2)


Note 1:
  If the voltage on a pin is within the specified range, the leakage
  current will be as specified.  If the voltage exceeds the specified
  range, excessive current may flow into the input; this is permissible
  iff the current remains below the specified maximum.

Note 2:
  Because RA4 does not have a diode to VDD, overvoltage may cause oxide
  breakdown.  RA4 will start sinking current somewhat below the breakdown
  volage, but the current required for device breakdown is very slight.
  External voltage limitting is thus strongly recommended.