> >The ability to use EEPROM or Flash parts for early development is a
> >fairly strong point, however a major customer will likely use an ICE
> >since the cost is only $1-5K and its usefullness is much higher than
> >Flash parts, on-chip debugging, and other glitzy technology.
>
> Sometimes (as in RF-related work) ICE's are not usable, and the only way to
> verify it in circuit is to have an adequate test plan in place ahead of time.
>
> I sense another thread starting....

Agreed wholeheartedly.  While ICE's can be nice at times, they are often
clunky and bothersome.  Even though my company has an emulator for some
of the CPU's we use, it seems I very rarely take it off the shelf: unless
I'm having a particularly vexing problem in my code, it's usually adequate
to simply use a normal CPU and reprogram as needed.  If I can ISP the
device, so much the better.

One thing I have sometimes thought would be nice would be a 24-pin PIC in
a skinnydip package with 18 of the pins in an ordinary DIP18-PIC configura-
tion.  This would allow use of a larger CPU during development/debug and a
smaller CPU drop-in once development was complete.  Unfortunately, I suspect
silicon probably wouldn't work out well for that and the idea would thus not
be practical, but it would be nice if it could work.