Origionally I was debating on this subject as well, but ended up decideing to use Normalaizing. I have it set to Normalize to 98% if peak is less than 90% or greater then 100%. This effectivly will only normalize if it's less then 90%....

The reason I decided to use it is because of the envirnoment of the car. If the cd track is recorded at a low volume on purpose, that is all fine and dandy for in the home or theatre or wherever, where the environment is much better and more suiting for the low track volume. In the car, often the road noise will make it to hard to hear the quiet tracks, so I normalize it to avoid having to adjust the volume when it plays, and then re-adjust after it is done for the next track. I find that most CD tracks have a peak around 98% as it is, so that is why I have it set there.

However, in addition to the above, I am only encoding these MP3's for use on the empeg, so obviously I am going to encode them the best to play on that and for the environment it is mainly used in (in my case, the Jeep)..
Just my two cents on the subject

--verteggio