QUOTE(KenH @ Aug 5 2006, 06:01 PM)

Surely the third option is most logical. Of course, you would probably need to use an image to define the boundary of the hair.
At first glance, absolutely. But once you're into the details, things change (note I do realistic characters but the following is true for any complex character model).
If you've worked with hair... or perhaps I should say attempted to create female hair-dos of any complexity, you would see the advantage to using emitters which are not part of the character model. Sometimes you need independent patchs that can point in any direction as well as take hair systems with guide splines that have different numbers of CPs or other differing attributes. Also, concerning the upper eye lashs and, to a lessor degree, for eye brows ; animating the model mesh so it looks good is rarely in sync with animating the hair emitters so the hair continues to look good and point in the correct direction.
Back to head hair -- beyond any desire to keep 'dos' and hair color independent of models so you can switch hair-dos among models which is perhaps something only I might need -- modeling the character head is rarely in sync with the proper type of mesh needed to emit evenly spread, problem free hair. Hair density is tied to patch size so evenly proportioned patches are desirable. 3 and 5 point patches as well as hooks can sometimes do strange things with hair. Also keeping the emitter patches separate gives much more control over the hair line (and you still use decals to control this). Trying to accomodate all these things to any degree while spinning the mesh for a character head is usually impossible or, at the very least, a major headache -- you completely avoid all this by making the hair emitters separate.
IMHO of course. :-)
Rusty
Edit: I'm surprised that 40% use the head mesh.