The DT MVD is meant to “transfer (pass on)” a design from one software to another one. If you want to do 100% full round tripping (e.g. have complete same editability between source and target) you are best to use the same software. That is a simple solution. You cannot expect Software A > MVD DTV > Software B > MVD DTV > Software A without any impact (although the suggestion from @Herb is appealing - that is: retaining as-is what you do not edit).
More than one way to geometrically describe an object
There is no single agreed way to describe the geometry of a wall. I immediately see several options:
- vertically extruded contour (e.g. IfcWallStandardCase)
- horizontally swept profile along a path (e.g. Complex Profile in ARCHICAD)
- side profile (sketch) with extrusion (e.g. in Revit)
- primitive box and Boolean Set operations (e.g. most CAD software)
- series of flat faces enclosing a volume (e.g. Mesh modellers, such as Blender or SketchUp)
- complex NURBS-based enclosures for organic shapes (e.g. Rhino)
- SubD Mesh models for an alternative approach to create organic shapes (e.g. Mesh modellers such as Blender or 3ds Max)
- Point cloud might even be another alternative
Modifiers
And you can then still combine this with all kinds of modifiers: voids, trims, cuts against other objects. Looking at the SEO (Solid Element Operations) in ARCHICAD also introduces some nice variants on Union, Subtract and Intersect which don’t fit regular CSG operations (e.g. subtract with upwards extrusion).
Semantic or geometric focus
Initially, Revit and ARCHICAD as “semantic BIM modellers” provided basic commands to model walls and called it the wall tool. More “Geometric modellers”, such as Rhino or SketchUp or BricsCAD give you the freedom to apply any sort of geometry you need. Gradually, the two approaches are getting closer to each other, where the user (designer) chooses a suitable geometrical modelling method which best represents the object at hand and then attaches meaning to it by classifying it as an “IfcWall”.
There is no single way to model a wall in any particular software these days. Why would you expect that the DTV would describe a single way to say what the geometry of a wall should be?
To me, it would be to describe the geometry as close as possible to the modelling method in the source application. If the source is a mesh, transfer a mesh (Tesselated BREP).