That's a tough question because every song or album requires a different approach. Of course, the perfect mix requires no mastering other than possibly bringing it up to whatever RMS/loudness you desire.
Allow me to digress a little on the subject of loudness of commercial material before I get into what I personally do.
The company I work for organises the largest pro audio/lighting/AV expo on the African continent called Mediatech. Part of the expo is a demonstration called the 'Outdoor Live Sound Demos'. All the local distributors of major brands like Alcons, L-Acoustics, DAS, JBL, Nexo, Martin Audio, etc put up their line array systems and we have a shoot out. These are all really high quality systems that you would find in any major music concert for thousands of people. Well, to cut a long story short, I was responsible for putting together the sample material for the shootouts. Basically I took three songs of different genres, loudness normalised them and then faded them into a 1 minute clip. I chose ACDC's Back in Black, a Michael Buble song (which I have always considered to be masterfully produced), and a David Guetta dance track (Titanium, I think). You can only guess which recording sounded the best on all of the systems. Yup, ACDC. I was actually appalled at how smashed even Michael Buble was. I made another montage with Snow Patrol and it was entirely unlistenable at the levels we were shooting for outdoors. Terrible. They really know how to ruin otherwise good sounding recordings with heavy handed mastering for loudness these days and it really showed.
ANYhoo. Just thought I would share that. It's easy to think that the top guys know what they're doing but when you hear a master recording of modern material on eleven or twelve superbly set up line arrays, it's easy to hear which ones are good and which ones fall apart. Most modern music fell apart.
In terms of my approach, I tend to first mix for the sound and the crest factor I'm looking for. To me, this is important. If I reach a comfortable crest factor for the material, i.e. a properly managed dynamic range, I find myself not having to limit the hell out of it in the mastering phase to reach the RMS values I'm looking for. This sounds like tricky business because a lot of guys on here, including myself, may tell me not to worry about that and just mix til it sounds good. I agree with that and that's fine but IMO a lot of what makes modern rock music (the music I generally record) pop is creative use of saturation and effect-driven compression. If you tier your dynamics processing in the mixing phase (and possibly the recording phase) by implementing multiple steps of gentle saturation and compression, you'll find that crest factor starts to come down without the macro dynamics of the track suffering.
Once my mix is done I'll use Wavelab to master. I actually can't claim to properly master because my room and monitoring system isn't transparent enough despite having a very high quality signal path. This is why I send serious work to a dedicated ME that I've worked with for a while. However, generally if I'm forced to self-master I'll use a BAX EQ and the Waves L2. I don't mess around with anything too complicated because all I want is bring the mix up in volume and adjust the top and low ends. I generally leave the midrange alone because I like to get it correct in the mix before it is ready to be mastered.
YMMV. Everyone has their own process and what works for them so take my words with a grain of salt.
Cheers