> Can someone explain in a nutshell what is the advantage to the Harvard
> architecture?  Or perhaps someone who's used other microcontrollers
> can give me some perspective on PIC advantages.  (Don't want to start
> a flame war, though, so relax, everyone.)

First of all, for the uninitiated, a Harvard Architecture machine is one in
which the program memory and data storage are entirely independent; both may
be accessed simultaneously without conflict and may have different word
sizes, etc.  A Von Neuman Architecture is one in which the memory space is
unified, allowing code to be run from data memory (and, if the code space is
in RAM rather than xxROM, data to be written to code memory).

If the code space is held in RAM, then Von Neuman machines can be easier to
work with than Harvard Architecture machines because the code may be easily
changed during development.  During production, however, these benefits are
often small.  Harvard Architectures may be less convenient to work with, but
since code and data are seperate it's possible to perform code fetches while
data is being processed (without requiring additional cycles).  This allows
code to run more quickly.

IMHO, for fixed-code applications, Harvard architectures are usually better
but for general-purpose computers Von Neuman machines are far superior.