Each situation really determines what you should do.
- Using generated coordinates gives you less control, but is easier.
- Using UV's is more precise, but harder, especially on complex meshes.
Now, I will use an example scene, to show some of the differences.
In this example, we will use:
- Sphere one uses generated texture coordinates
- Sphere two uses sphere projection, to create a UV map.
In this circumstance, I like the results of sphere 1 better than that of sphere 2.
When using generated texture coordinates, keep in mind what mode of texture projection you are using. I have a sphere, so I am using sphere projection. For other objects with complex meshes that you do not want to unwrap, box projection is usually a good option.
When looking at the UV map that was created through sphere projection on sphere two, you may notice that the unwrap is not perfect, and that is one of the advantages of using generated texture coordinates.
Also, remember that using UV or generated texture coordinates are not the only options. Using object coordinates can come in useful when you want the texture's projection to change based on the position of the object, such as when creating roof tiles. You may use a cracked texture, and want the cracks to be in a different place per tile.
So when deciding, just think about what your goals are, and what the easiest way to get there is. In the earth model case, I would use generated coordinates with sphere texture projection.