This is currently one of the biggest issues in mastering...
Now that music can be squashed and EQ'd to a continuous maximum volume (The Loudness Wars), we're all faced with doing this (or not) to our material... Originally mastering was actually the process of creating the transfer process from tape to vinyl with vinyl's limitations... "Sweetening" was applied to try to compensate for the loss (Read about RIAA curve)...
In it's current manifestation, tho sometimes it can be to correct for mix problems and the likes... It's main purpose seems to be to increase gain... That comes at a cost of dynamics/tone and deciding such importance is much a matter of taste... tho when you hear your material broadcast back2back with material mastered loud, it can feel bad when not near as loud... If someone liked what they hear they will probably turn it up and not think about it...
I'm currently in my own dilemma with this too... tho I'm a seasoned recording and mix engineer, I'm not actually a mastering engineer and I seriously question every piece of work I do... I'm finishing my 8th album right now and I've mastered it two ways... The first is a very minimalist approach... Nothing more than a tiny bit of peak limiting... The second is a more "modern" approach and uses a number of levels of compression/limiting/etc... and it is louder, but I don't know if I like it better or if I'm simply impressed at the relative vlume when played against other peoples' work...
Hard to decide...
But there's somewhat of an answer for you... It's not actually the mix, it's how much of this do you apply to your mix to increase the volume and can you still enjoy the listen! ;-)
Best of luck... I still don't know which mastering approach I will finally use...
I wish I could hire a mastering engineer as well seasoned as I am in my work... but alas, I'm also worse than flat broke! <sigh>
Keni