That question has been answered multiple times already, it would be very nice to stop running in circles. Native media source is a simple demuxer, it NEVER outputs decompressed formats.
This, !5615, and other later changes, will improve the code the following way:
1) Save 1500 lines of code from winegstreamer, by removing the need for an intermediate representation and the translations from/to it to various other formats.
2) Improve compatibility with a lot of games and applications, by implementing actual native behavior of the media source, and letting MF pipelines be resolved as they are supposed to.
3) Improve compatibility with pretty much every Unreal Engine / Unity game by enabling D3D-aware pipelines, which many games expect and assume to be working in many specific ways.
4) Improve performance by using zero-copy, as it is supported in decoder and video processor, and eventually using D3D for video processing instead of software color conversion.
5) Use standard translation functions to convert between formats, which can be tested, as opposed to that intermediate representation which cannot.
6) Allow arbitrarily long codec data, which is currently capped to some arbitrary limit, without any manual tweaking required, as well as easier mapping of any custom codec format fields, if we ever need to do it without having to update the custom representation and every mappings.
7) Still, allow any compressed format to be represented and decoded, avoiding any regression with the current code, while fixing every new format games might be using, without any manual intervention or additional code being required.
Now, instead of asking the same questions over and over, could you give me some actual reasons why all this is not desired?