metric and measure on the projective space Let $RP^n$ be the $n$-dim real projective space. I have the following four questions.


*

*What is the so called standard metric on $RP^n$? 

*More generally, consider a metric space $M$ with an equivalent relation ~. Then there is a natural way to define a (pseudo)metric on the quotient space $M$\~. See Why are quotient metric spaces defined this way? for details. It seems to me that if we regard $RP^n$ as the quotient from $R^n$\{0}, then the induced pseudo metric is not a metric, however, if we regard $RP^n$ as the quotient from $S^n$, then the induced pseudo metric is a metric. Am I correct? Is there a contradiction?

*What is the standard measure on $RP^n$?

*How to in general define a measure on the quotient of a measure space? (similar to question 2)


Thank you for reading and thinking for my long questions. I put them in a logical order, however, I'm mostly interested in the third question. If you can't help me with all the questions, please try Q3 first.
 A: *

*Apparently you ask about metric as two-point function, not about Riemannian metric in the projective space. The answer  (consistent with the Riemannian metric) is: angle between lines through the origin in $\mathbb R^{n+1}$ (which represent points in $\mathbb RP^n$). The diameter of the space is $\pi/2$. 

*Correct on both counts. Using the Euclidean metric on $\mathbb R^{n+1}\setminus \{0\}$ and collapsing along lines leads to a disaster: the quotient pseudo-metric is identically zero, because the Euclidean distance between any two lines through the origin is zero.  This issue does not arise with $S^n/\{\pm I\}$, whether we use the restriction (chord) metric on $S^n$  or the intrinsic (arc) metric. The latter choice is more natural, and yields the metric described in 1. 

*Pushing things from $S^n$ to $S^n/\{\pm I\}$ by the quotient map is a convenient and natural way to equip $\mathbb RP^n$ with various structures, including the measure. The pushforward of normalized Lebesgue measure on $S^n$ gives the standard probability measure on  $\mathbb RP^n$. 

*First, try pushing forward by the quotient map. Sometimes this does not work: for example, pushing the Lebesgue measure by the quotient map $\mathbb R\mapsto \mathbb R/\mathbb Z$ we get something ugly (which gives infinite measure to every nontrivial arc of the circle). In such a case, consider restricting the measure to a fundamental domain first.
