Well, everyone can only express their own opinions based on their own experience and the path they chose.
If you really want to learn it and not achieve some immediate goal with finishing your planet rendering program, then my opinion is: start learning modern OpenGL.
I myself started learning the legacy/old fixed-function pipeline OpenGL in 2007 and transitioned to more "modern" OpenGL 2.0 (VBOs and shaders) two years later. This has been at the expense of many many maaaaaaaany hours of googling the web left and right and up and down for OpenGL-related resources, articles, papers, specifications, forum posts, blog articles, literally everything, building dozens of little test and demo programs and throwing them away after reading how that all can be done better. If you get hooked by the thrill of learning it, and the small moments when it makes "click" in your head, then there's no stopping you anymore.
The hardest thing but also the most fun is to get an accurate mental model of the object model, state management and the rendering pipeline in modern OpenGL.
In modern OpenGL you just have a few core concepts/objects that you use for everything (VAO, buffer objects, textures, samplers, shaders and framebuffer objects). Then there are the different shader stages doing different things. Sooo many things to explore.
So, start digging around and building your own mental model of modern OpenGL.
I guess, at the beginning literally every starting point is a good point.