Hardware SPI+SPI or SPI + Software I2C?

I'm trying to decide which is a better approach. I'm using a 16F877A and either a SPI based UART (MAX3110E) or Ethernet (ENC28J60) IC.

I'd also like to use an EEPROM, Either a 25lc256 (SPI) or a 24LC256 (I2C). The EEPROMs don't need to be fast an could be hardware (SPI only) or software (I2C or SPI). The hardware solution requires a CS line for the Chipselect.

Problem I2C appears to have much larger EEPROM options and I could keep the SPI bus free for use with ENJ28J60.

Opinions?

well for me, it is much faster to develop software SPI routine.

On the AVR chips, the SCLK frequency can run at up to 1/2 the CPU clock, which makes the software much easier to write and debug (and scope out when something goes south). Plus it being a dedicated full-duplex connection means that the routines aren't nearly as complex as the whole I2C state mess.

Admittedly they eat up more pins, but when you use the chip select lines, it's pretty easy to abuse/multiplex more functions on to the SPI pins.

Problem I2C appears to have much larger EEPROM options

Click to expand...

Are you referring to capacity? Both Atmel and ST have some pretty huge SPI flash parts... (unfortunately I don't think they're compatible in any way)

James