Bump up to a '1284 tho and you'll have access to the most SRAM of any arduino - 16KB.
'328 doesn't give you 2 hardware UARTs. Or JTAG access. 32 IO without going to external hardware can be a benefit too - 16 bit shift register is only 1 direction, need a 2nd one to be bidirectional. Andwhat if you're after a couple more analog inputs?
I can even sell you a board for $5 (mailed to US location) to build up into a duemilanove style board.
Populate as you see fit from your parts bin and with connectors to suit your project, with an onboard mikroe483 FTDI module ($11 at mouser.com) or plug in an offboard FDTI Basic type part for sketch downloading only.