It really depends on the genre of music.
I'll usually put a bit of compression on bass, guitars, and vox in punk and metal stuff, just hitting the peaks and smoothing the signal out a little.
For Blues and Jazz I like to give everything more breathing room.

I'd rather ride the faders than squash the tracks.
If I do grab a comp it's for a little extra control on the vox, bass or guitars.
If an instrument is getting way out of hand I'll stop recording and call a do-over, sometimes that includes a discussion with the musician
about his or her abilities and technique.

As far as the final product entering the loudness war, I'll let previous posts here answer that one for you.

Well, that's pretty much the way i do it too, but a couple of days ago, i loaded a song (from a commercial CD) in a ProTools, and i was amazed byt the lack of dynamic of it.
I mean, i never saw that before, and i was kind of wandering how they could achieve such compression/limitation without ruining mixes... Or is it that the mixes are already hardly squashed down???
If i try to take one of my mix and make it as "loud" as this one, well, the result is far, FAR from the mix in the first place...
Do the mixers foresee that or is it just a mastering process to turn those thing that way??

Its a bit of all of the above.

Mix engineers are compressing their mixes more, and mastering engineers are compressing them EVEN more. Dynamics have gone the way of the dodo bird over the last 5-10 years. But lately the phenomenom has been creeping well beyond just mainstream music too. I just picked up Sondre Lerche's new album "Phantom Punch" and was really surprised by how squashed it is. Usually his albums have had a lot of dynamics and energy.

I just don't get how they can achieve so little dynamical product and still havin it sound, well, not good, but but considering the ration of distortion, that's amazing you can still here somethin'.

I mean, how can a snare drum still sound like a snare drum when there is little romm for it??? I mean, technically, hox can they achieve that??? Even if that's bad and everything, there must be some tricks here...!!! Ain't there??

Does it sometime makes sense to over compress a mix, saying like in heavy metal or somethin??? Or do they foresee that it's gonna be done afterwards and then they'd rather control it, at least a little bit?? Or do they really just want to achieve maximum loudness?? Cause i've never been able to listen to that kind of music without extreme compression.

I remember the salesman from Cedar tellin me that he often sells cambridge station t mastering facilities so they could declik the artifacts due to extreme limiting... Ain't that crazy!!!

Anyway i've been doin a lot of readin lately, and i must say that i am pretty relieved to see that everybody is starting to see that as a problem, and tryin to go back to putting music first...
I hope that i'm not wrong by saying that, i think the revolution might soon be on...!!!

Workin in sound restoration, i would compare that problem to extreme denoising in our industry...
Some clients just want less and less noise left, even though you have to hurt the voices and stuff...
We oftenly have to seat our clients in the mixing room and talk to them and make them really hear the difference... And hopefulle, in 99,9% of the case they realize that extreme denoising is really not the way to go...!!!

But it's more easy, at least for me, to understand how extreme denoising can be performed than to understand, even if I know how about limiting and everything, how you could still have sound, even if it's bad, produced when you are workin in a 0,5 dB dynamic.