I used something similar to that decoding scheme in my first 6502 build. The R/W qualification with PHI2 is to ensure that the address and data lines have settled and the previously accessed device has released both buses - this avoids bus contention. 65C22 and 65C51 devices take this in to account themslves and so don't need to be qualified with PHI2.
With regard to the /OE lines: I think the circuit will set up all attached ICs as if they're going to be accessed (which includes R/W and /OE lines), but only select the chip that is decoded using /CE. I think the /CE input on the chip changes faster than the other inputs as well so it does make sense to do this to get the selected chip to respond faster - the setup is already in place.
I've included an attachment with the decoding that I used and the RAM IC - as you can see, the RAM is the only device which is qualified with PHI2 (CLK in my diagram). The VIA (65C22) doesn't need it as it takes into account the need to act on PHI2 going high, the ROM is slow and the contents cannot be corrupted as well. This leaves the RAM (normally much faster than ROM and can be corrupted if the busses aren't completely set up).
I hope this helps.
P.S. please note that the decoding scheme came from Garth Wilson's primer as well (see
http://wilsonminesco.com/6502primer/potpourri.html), although I'm now moving on to using GALs.
[edited to add more info]