I don't know if PDF has truly evolved from its desktop publishing origins, but it is a terrible format because it no longer contains the higher level source information that you would have in an InDesign or a LaTeX file. PDF/Postscript were meant to represent optical fidelity and thus are too low-level abstractions for a lot of end-user, word processing tasks (such as detecting layout features), and thus trying to reverse engineer the "design intent" from them feels like doing work that is unecessarily tedious. But that's the way it seems to be given the popularity of the format.