I'm reading Milnor's book on Characteristic Classes. In Appendix C, Milnor shows the invariant polynomial of the curvature form and the Chern class differ by powers of $2\pi i$. He first shows that the first Chern class and the trace of the curvature form are multiples of each other by a constant $a$. He calculates this constant for a bundle on a sphere, and finds it is $2\pi i$. Milnor says it's enough to calculate $a$ for one specific case, and that this constant is the same for all vector bundles.
My question is: why is it enough to find $a$ for one specific case? Is there any reason why the constant can't be different for different bundles?
The proof is done on page 306.
The point of the proof is to show that what Milnor calls $\Omega_{12}(M)$ is a characteristic class for complex line bundles $\zeta\to M^2$. It's not an arbitrary element of $H^*(M, \mathbb{C})$; it's a map from bundles to cohomology that's natural and defined in terms of a certain classifying space. On p. 298, Milnor proves that certain forms induce characteristic classes; he then proves on p 306 that $\Omega_{12}(M)$ is a form of that type and thus defines a characteristic class. The only characteristic classes for complex line bundles are of the form $\zeta \to \alpha c(\zeta)\in H^*(X, \mathbb{C}) = H^*(X)\otimes \mathbb{C}$ for some fixed $\alpha\in\mathbb{C}$ (see, for example, the axiomatic definition of the Chern class--- it's determined by its value on the tautological bundle of $\mathbb{CP}^\infty$), and going through the computation for the case of $M = S^2, \zeta = T^*S^2$ determines $\alpha$.