> On the other hand, the markdown image is hardcoded as a special case in the parser, while the rst image is not. It was added in the exact same way as every other directive in rst: register a handler for the directive, have the handler output a specific kind of node, and then register a renderer for that node for each builder you want.

Why not "hardcode" the most common things to be the easiest to use and then still have the option to extend to other protocols? Why "suffer" every time equally instead?