Hi Samy,
As a non-Rigger, I'd say neither and both and it depends...
If it were as simple as a rule, or personal preference, everybody could be a Rigger.
But I'll let Wayne (or someone else with more Rigging knowledge than me) answer this, he (or they) will have a better, more nuanced answer, for sure 😉
Constraints are chosen based on desired effect of the rig. For example: if you want a single control bone to set where the eyes of a character is looking at. You would use a tracking constraint on each eye pointing to a bone. Then these bones would be parented to the control bone.
In some cases it's preferences or "That's how I learned to do it." An example is parenting bones. You can do it thru normal parenting or child if constraint. The general rule is If parenting isn't going to change then use normal parenting. Note: there is also the armature constraint that can be used for parent switching too, but that is more advanced rigging.
Hi Samy,
The constraints are generally chosen by the effect they have, as Dwayne mentioned above.
Experience will inform your decision making - but often there is more than 1 way to achieve the desired outcome.
In the end it's about solving a problem.
"I need the rig to do X", and then you figure out the puzzle so the rig does X.
There is a lot of higher order thinking involved initially, however, with a lot of the things you practice and repeat - you start to see patterns and then don't have to resolve problems....because you can see it's actually the same problem - so the same solution will apply.
Hope that helps.