In article <Pine.NEB.4.51.0305212314380.19204@paper.durnsford.net>,
Gavan Fantom  <gavan@coolfactor.org> wrote:
> A minimal PIC programmer shouldn't cost you more than a few dollars if you
> build it yourself. I use a setup where the main cost is the parallel port

Sure, but I'd rather spend the few bucks and get a system that

a) I know uses the right voltages for the right parts at the right times, etc
b) lets me do in-circuit debugging with the windows software if need be
c) lets me concentrate on developing my new thing instead of reinventing
   something that's already available off-the-shelf at an OK price :-)

But, if part of the reason you're doing it is for learning and fun, and
you have the time to do your own thing, that's cool too. Personally my
time is quite limited, so I chose to buy things that will save me time
and give me more time to work on the interesting parts of the problem to me.

> Also, for the application in question, cost savings could be made in
> hardware by using a cheaper PIC and implementing a software UART. There

The price differences between pics are at most a coupla-three bucks. 
Like, IIRC qty 25 of 877A is $4.25 or something.  The absolute low-end
cheapest PIC in that qty is $.97 (the tiny 8-pin OTP parts).  So,
there's a limit to how much you can save :-).  An order of magnitude
less savings choosing between pics than between pics and BS2's, anyway. 

Also, I just recommend the 16F877(A) because it's easy to devlop there,
and then move it down to whatever PIC is the smallest fit, if you want. 
Just because it has a UART doesn't preclude you from bit-banging serial
if you want to or think you might want to move it down to a uartless pic
when you're done :-)

Of course, there's also the no-MCU route too, but the MCU gives you a
lot of flexibility, and PICs are so cheap that it may even cost less than
using some task-specific part.

Anyway, there's a while range of options with different tradeoffs... 
choose what makes sense based on whatever factors are most important to
you.