With the UML, there is such a thing as a structured classifier. Structured classifiers allow for the modeling of the constituent parts that the make up the classifier.
This is handy because we can then show internal parts within the classifier, the connections between the internal parts and the boundary of the classifier itself. These connections can also illuminate the intricacies of the interface connections amongst the internal parts of the classifier.
So your main choices of modeling elements to use when showing a structured classifier are components and structured classes. I think it makes much better sense to use a component for this purpose. Components are geared slightly more toward the logical structure of a classifier and a a component is a structured classifier. Structured classes allow the same thing but I think classes are better used in the traditional way they were used before the advent of structured classifiers. The are good at showing the attributes, operations, and relationships with other classes. Components can show the same class in a different diagram but from a structural viewpoint.