I must apologize to physics forums I drink a lot
I'll try to answer your question about fiber bundles and non-abelian gauge theories in one fatal swoop, in a physics way, not to mathy.
Ok, you have a manifold, M. At each point on the manifold you can create a vector space of all vectors tangent to the manifold at that point. This is called the tangent vector space at that point, TM_x. Now you can think of a disjoint union of all the tangent spaces at every point on the manifold and this is called the tangent bundle <br />
TM = \coprod\limits_{x \in M} {TM_x }. Similarly, the cotangent bundle is the disjoint union of all the, orthogonal vectors, one forms to each point on the manifold <br />
T^* M = \coprod\limits_{x \in M} {T^* M_x }. The important part to a physicist is that the directional derivative of tensors changes from point to point on a manifold. The connection A or in relativity \Gamma _{\beta \gamma }^\alpha tells you how the directional derivative changes from point to point on a manifold. This is where the concept of a bundle comes in. So what you get is an exterior derivative. For Yang Mills, non abelian gauge theories, this exterior derivative is D_\mu = \partial _\mu - iA_\mu ^\beta (x)t_\beta where t_\beta are the generators of semi simple lie algebras. Now, looking at the holonomies, parallel transports, of the exterior derivative you get P\exp \left({i\oint\limits_C{d{\mathbf{x}}{\mathbf{A}}(x)} } \right). It should be noted that this is very similar to looking at the fundamental groups in topology. Except in that case, you let your loops get contracted to the base point. Now expanding P\exp \left({i\oint\limits_C{d{\mathbf{x}}{\mathbf{A}}(x)} } \right) you get <br />
P\exp \left( {i\oint\limits_C {d{\mathbf{x}}{\mathbf{A}}(x)} } \right) = e + \frac{1}<br />
{2}\iint {dx^\mu } \wedge dx^\tau \left( {\partial _\mu {\mathbf{A}}_\tau - \partial _\tau {\mathbf{A}}_\mu - [{\mathbf{A}}_\mu ,{\mathbf{{\rm A}}}_\tau ]} \right)<br />
+ \cdots
. Now the curvature term is {\mathbf{F}}_{\mu \tau } = \partial _\mu {\mathbf{A}}_\tau - \partial _\tau {\mathbf{A}}_\mu - [{\mathbf{A}}_\mu ,{\mathbf{A}}_\tau ]. What they mean by non-abelian is that [{\mathbf{A}}_\mu ,{\mathbf{A}}_\tau ] \ne 0