For most object in Blender rotation animation curves deal with Euler angles. However when dealing with pose channels in an armature, the animation curves for bones deal with quaternions.
How to people feel about having to work with quaternion curves for refining armature bone rotations versus the Euler curves that apply to other types of objects?
Doesn’t bother me. I’m usually (trying) to do character animation, and I really don’t need to see any numbers other than the frame number, (probably don’t even need that. The general shape of the curve vs time is enough to be able to adjust the timing. As far as which curve affects what rotation, I usually just ctrl-LMB a test point and move it around and see what the bone does … then adjust from there 
Here’s a really great explantion of how they work for anyone else (like me) that doesn’t really understand them
Mike
How to people feel about having to work with quaternion curves for refining armature bone rotations versus the Euler curves that apply to other types of objects?
Do we have a choice?
%<
I understand that quaternion rotations avoid Gimbal Lock than can occur with Eulars. If this is so, then I guess I like them.
Frankly, as Mike said, I just grab a control point and see what it affects and go from there. Also, as Fligh said, what choice do we have anyway.
Unlike object rotations, that may be calculated differently than you expect between key frames, I rarely find any problems with bone rotations.
The choice we have is that a request can always be made to the Blender team to try to implement something else.
I’ve been having problems working with quaternion curves, so I’m trying to decide whether it is me or if quaternion curves aren’t the best way to be dealing with animations.
Lots of other objects in Blender can be animated with rotations, and they don’t use quaternions. Also, I’m not aware of any other 3D application that exposes a quaternion interface for rotating anything.
Something for you to try, that I’ve just been experimenting with.
Add a copy rotation constraint to the bone(s) you want to rotate with an “ordinary” (non-armature) object as the target. You can then rotate the “master” object, and the IPO window shows Euler degrees (I think that’s what they are. They’re not (displayed as) quats anyway and they correspond to the N(umerical) input box. You can change the tranform orientation of the master object (local,global,normal) and get different results/effects.
This is really handy if you use it with the “record mouse” technique as described this thread
Mike
I do a lot of animating, and I don’t get why you’d want to fool around with the quat curves themselves, other than perhaps changing interpolation modes or pulling handles to tweak ease-in and ease-out.
I’ve always found it quite handy to just put the bones where I want them and set a key. If I don’t like it, I reposition the bone and reset the key.
Not saying you don’t have a use for it, but I just don’t see it. Having talked with Ton about quaternion vs. euler issues, I can tell you that you’ll probably never be seeing adjustable euler curves for bones.