Mastering in the Mix

In today's computer music production environment, if oneself is doing both the mixing and mastering, is it necessary to export the mix as an audio file and then open a new project importing the audio file to do the mastering? Why not master within the mix. Personally, I mix with all sounds rendered to audio tracks. Midi tracks are archived and hidden away; instrument tracks deleted; no freezing. And the buffer size is set way high for the audio device. What do the audio experts in this forum say?

Depends on the quality , for demos and just for fun songs whatever get's it out there, but for those hit records I think you'll still want to master. 
I just find it much easier to get my mix right , export, and then top and tail and make fine tuning to RMS level etc  in a wave editor. Sonar is my multi track, Wave lab my 2 track. 

That's why I like T-racks mastering. I master outside of sonar but the file is still linked to the sonar export so if I need to change something in the mix I can without having to reload and start the master process over. It just update my changes automatically. Can't wait until Halloween for version 5. Cheers :)

One of the best reason I can think of to do the traditional steps (track/mix/master)is that way your compression, esp., is incremental and serial.
 
A little compression coming in during tracking.  Even if you only shave a few dBs off the loud spots during recording it makes each track look a little more compact - ie. there are smaller overs the average level.  Which means your track is a little louder than it would be, on average, and isn't distractingly dynamic for mixing purposes.
 
When you mix these preprocessed tracks together, you can further restrain the louder sections w/  gentle compression on the track - or heavy if that is the sound you want.  Gentle is the key here for louder music as the end product.  Again, you are shaving a few dB off many tracks.  If you look at the audio in your DAW these tracks aren't spikey with deep troughs, although they don't look ruler flat either.  They still have dynamics, just not as wide.  They are, not surprisingly, compressed between the loudest parts and the quietist, so there are now only a few dBs difference between the two.  If you have already used level automation, or will, the compressed track now slides into the song and should maintain a fairly steady level within the song.  Example, your rhythm guitar doesn't disappear so much in the soundscape as other instruments come in and dominate.  It doesn't pop out either with that one hard hit note, but you can still easily find the guitar part it if you listen.  It is solid.
 
Why not just slam the rhythm guitar and put it exactly where you want it in the mix?  One, when coming in, 1176 style compression would often be used - it is like an on/off switch it is so fast.  Perfect for a chunky rhythm guitar part.  But the 1176 imparts some saturation or edge to a sound, with more edge the harder you hit the unit.  Hit it hard during mixing, and any more saturation during mixing can cloud the track.  It is hard to close a large gap between the greatest and average sound levels in a track at one go, not impossible, but fraught with unintended side effects.  To me, it sounds more natural to go with layers of compression and there is less chance of damaging the track.  Now your mix is, well, mixing a bunch of tracks that are constrained in their dynamic range - not dead dynamics, but simply smoothed out during tracking and mixing.  They persist solid in their soundscape slot and let the lead vocals and lead instruments fill their assigned spot in the soundscape without losing the groove.  You've got a solid mix. 
 
Mastering is just the final polish.  Subtractive EQ can also reduce any frequencies that jump out, thereby lowering the AVERAGE volume level, and rolling off the bass so you can, again, raise the AVERAGE level of your song with make up gain  And more gentle compression to squeeze the differential between the quietest and loudest  part of your song.  Now your song can sound loud throughout, and any breaks with buildups or quiet bridges still sound loud and proud enough and not discordant with big volume changes that can happen with everything compressed once.
 
Tracking gets the best sound from the source  possible within the demands of the song itself.  Mixing gets the most overall volume between tracks of the song, while mastering lets you concentrate on getting the most volume on the song itself, and between other songs.  To get there the easiest and simplest method is serial compression at each stage where appopriate, and until you figure everything else out the surest method of getting volume without damaging your tracks.