If you don’t want to import color or texture information, there would be the possibility to give each separately animated part of the model its own color, then read the color information from the 3d file and use all polygons/vertices of the same color to determine animated object parts.
It pretty much depends on what type of file you’re importing though, many 3d object files are structured for having ‘sub-objects’.
Another possibility might be, to build your objects as a scene file (each separate animated part being one object in the modeler) and importing the scene file.
Quake 1 and 2 use simply a series of animation phases… first comes a cloud of 3d points, then the definition which vertices span which triangle and then texture coordinates. This is repeated for the number of animation phases in the model. IMHO not in Quake 1 and 2, but 3, they interpolate between the animation phases to make the animation look smoother (keyframe interpolation).