Hmm.  Is the failure of circuits due to use of the internal clamping diodes
due to some internal issue, or is it due to "large" currents raising or
Lowering Vcc/Gnd out of spec?  Given a relatively high impedence power
source (say, lithium coin cells), it's easy to imagine "clamping currents"
on the order of what the PIC itself draws causing all sorts of Vcc related
problems.  (I have joined the conversation late, but I can't see an
electrical reason why the internal clamping diodes would behave much
differently than external diodes in a similar configuration.)

This is an important question - if there are Vcc problems as opposed to
internal issues, that means that your external clamping circuit has
limitations as well - in particular, the external schottky diode scheme
Russell suggested isn't a good idea.

(In particular, someone mentioned an "rs232" device.  If that circuit were
powered by the rs232 port as well as using the internal diodes to clamp the
signals, that would mean that the power suppy AND the signal had (about)
the SAME impedence, which does sound like a recipe for disaster!)


    I wouldn't attempt to interface the Inputs and Outputs of ANY
    micro-controller or microprocessor directly to the outside world.  ...
    The very very minute cost of either diodes, transistors, or buffers
    ICs is just not worth destroying your project.

You're making the old-fashioned assumption that the cost of the protection
circuitry IS "minute" compared to the rest of the circuit.  In these days
of $0.50 microcontrollers, that's an unjustified assumption!
Semiconductors is semiconductors - treating the microcontroller as though
its some sort of sacred cow is ... outdated.

(Just trying to have the ON-TOPIC flames keep up with the off-topic
flames.)

BillW