I did not intend to replace the POF format. I want to implement the glTF format as an alternative. For this, I've changed the thread's title.
Doesn't matter, it's still a stupid idea that is way more complicated to implement than the better alternative.
The basic problem you're trying to solve does not exist in a form that can be helped by incorporating support for non-POF model formats
into the engine. On the contrary, it creates a whole ****ton of issues that
you, in my considered opinion as a professional software developer, are not able to solve.
FSO, due to its lineage, makes certain assumptions. One of them is that models will always be in pof format, which means that everything that interacts with model data is optimized around pof, its features and limitations. In order to incorporate a new model format as seamlessly as possible, the engine would have to perform an on-the-fly conversion of that model file to POF - and that leaves out some issues that are created by the simple fact that other model formats
do not contain the same metadata that pof files do. Subsystem data, firepoints, moment of inertia data, all of that needs to be supplied
somehow, after all.
Now, that's not an insurmountable issue per se.
But: If you're doing all of these things in order to end up with a data structure that is indistinguishable from a POF, why aren't you just using a pof?
In other words, support for new model formats is great. But the proper place for it is in a POF authoring tool, not the engine.