First let R and S be dvrs, and let t be any generator of Q. If the valuation of p is 1, then the valuation of t is 1/d. This holds for every conjugate of t. Assume the extension is separable, so that the norm of t is a product of conjugates. Each conjugate has the same valuation. Multiply d conjugates together and the valuation becomes 1. The result is p times some unit in R. This happens c times over, for all possible embeddings of S into the algebraic closure of E. Therefore the norm of t is pc times some unit in R. This holds for every t, so let |Q| = the associate class of pc, or if you prefer, Pc.
In general, norm and localization should commute. (They do when taking the norm of an element.) Localize about Q, and both rings become dvrs. So |Q|, localized about P, yields Pc. Localize about any other prime and Q becomes the whole ring. Represent this by 1, or any other unit (having valuation 0), and take the norm, and find a unit in R. Thus |Q|, localized about any other prime, is R. The factorization of |Q|, in the dedekind domain R, has to be Pc.
We can say that norm and product commute by definition, and that's ok, but we would like this to be consistent with the traditional definition of norm. Assume x generates a principal ideal that is a product of prime ideals, that may not themselves be principal. Now we have the norm of x (in the traditional sense), and the norm of the ideal {x} generated by x. The key observation is that these definitions agree locally. Write {x} as a product of prime ideals, localize about any prime ideal (thus giving a dvr), represent ideals with elements, recall that norm and product commute when dealing with elements, and establish equality between the two definitions locally. Since the norm of x commutes with localization, and the norm of {x} commutes with localization by definition, pull back to find the norms must be the same globally. All is well.
This definition extends to fractional ideals in a natural way. The norm of Q-1 is P-c, and so on.
There are more general ways to define the norm of an ideal, but I'm not going to get into them here.