gEDA-user: Multiple footprint methodology?

[Dave N6NZ]

Hello,

Total gEDA noobie here, but my first encounter with EDA workstations was 
about 1983, so I know the vocabulary and the work flow. But I have a 
looooong learning curve to climb in terms of "how gEDA does it".

This is probably a FAQ, but after searching the archives I couldn't see 
anything that was relevant and up to date.

My specific question:

Given a part with two or more packages, how should the footprint mapping 
be defined in the symbol? My specific example: ATMega88 has three: PDIP, 
TQFP, and MLF.  It is desirable to be able to simply flip an attribute 
to go from a through-hole board to surface mount, so it seems to me you 
want the same schem symbol to be net-listable either way with no other 
changes to the schematic other than an attribute change.

Related question:

In the case of the ATMega88, the TQFP-32 version of the part brings out 
more I/O than the PDIP-28 version. Soo... choices. 1) Define all the 
pins, but don't map them to PDIP-28?  Will net lister issue warnings? 2) 
different symbol?  Icky, because you would like to switch packages 
without deleting/inserting a symbol.

Another related question:

ATMega88 is actually one of a family of 3 parts: ATMega48, ATMega88, 
ATMega168 which have identical pin-outs but different sizes of internal 
memory. Again, if you have a board designed around an ATMega48 and go 
"oooooops -- code bloat -- need an ATMega88", then it would be nice to 
be able to flip an attribute and get a new BOM.

TIA,
   Dave