There is a big difference betwene 2 buss/master compression and individual channel compresison. They are often used for two completely different reasons and one does not really negate the need or desire for the other. In general, when speaking of 2 buss and master compression I tend to gravitate towards the compress first camp. A big part of the reason that I do this is because EQ'ing first will change the way the compressor triggers and as a result will change the output. This could be a good thing, but generally I prefer it to not be this way. The advantage would be the already compressed output which would help contain peaks, but then that could also be viewed as a disadvantage. One other reason I like EQ last is that it seems to give add back the appearance of dynamics to a compressed source without completely negating the compression itself. Pretty much I do it the way Glenn described above, except that "fix EQ" as he puts it is where I typically address things at the individual channel layer, or maybe one of the group stems.