I'm going to go against the grain here and say that I very much prefer to normalize my tracks before mixing. The reason being that then the faders in the actually represent the relative volume levels of the various tracks. Without normalization if you have something from a quieter source mixed with something from a louder source it means your faders may have to be at drastically different levels to get them to same relative volume. Once the tracks are normalized the volume faders can be used to set the desired perceived volume levels, and looking at the mixer will actually give you some useful representation of the relative levels of your tracks. Otherwise you can get the situation where track A is much louder than track B and yet its fader is at the same or lower level.
Since this is still all internal to the DAW, there's no danger of clipping the output, it just makes levels more sane to reason about. I use Reaper as my audio tracking DAW and it's flexible enough to allow you to normalize individual clips, and then even tweak the gain of each clip after the fact.
I never normalize a final mix though, and just determine the final volume based on material.
I'm by no means an expert at mixing, but this approach has worked pretty well for me. Better than not normalizing anyway.Just so it's clear. Assuming two tracks, A and B