Dozer! It should make your job so much easier to be able to just switch between RMS or Peak and decide based on texture. We used to have to press loads of buttons or keep changing and fumbling with patch cords.

I like lots of both and plenty of neither. They both have their applications. There is no correct choice. Sometimes I would cut vocals with my DBX 165A's, sometimes on the UREI 1176's. I would strictly choose based upon an emotional level, or if I didn't like what I was getting by twisting the knobs back-and-forth. The 1176 had no automatic settings whereas the 165A had both automatic and manual settings and the ability to choose either hard knee or soft knee. Imaginative if boyfriends were like that?? I'm getting dizzy.

If I wanted something under tight control, it was usually the 1176 with its peak sensor, fast attack and response. If I needed to smooth out some keyboards, frequently I would grab at the 166's or 165A's strapped for stereo. And let whatever peaks there are to come through unless they were a banger. In which case, they were limited beyond dynamic comprehension. With nary a peak.

If I had the lead vocalist on the 1176, I would frequently put the backup vocals on the 165A or ,sometimes vice versa. Once I was requested by an FM radio station to downstream a stereo limiter before their real-time MP3 coder. For that, I didn't want RMS. So I used my Orban 418A stereo limiter. Which was a peak sensor limiter, originally from the FM Opti-mod design. I was able to adjust the release time for the amount of "apparent loudness" and density, I wanted to impart to the mix. It worked out well even though I don't like mixing through limiters. It was for the Counting Crows, even though I lost count.

Densely limited beyond limitations
Ms. Remy Ann David

Or... to put it another way, start with this rule of thumb:
The more dynamic the input, use peak,
Less dynamic use RMS.

So for example: If the vocalist has a lot of sibilance like on T's , S's and P's or hits a powerfull note that send the meters flying, use Peak. (actually there's other tools for that, but that's another topic) If the singer is just louder in the chorus and too soft in the verse, use RMS.

Understand that these are just starting points. Like Remy said, from there it's a matter of taste and attitude.

I enjoyed the "natural" response as I had wondered that myself. Word up!

Hey Remy, do you happen to remember the date of that broadcast, it's extremely likely that i have it :D
not that i'm a bootleg collecting junkie or anything
and i have hundreds of Counting Crows boots