I have tried to describe what the current MVCSource plugin does
here.
To summarize, the decoder should ideally be able to take the two demuxed video streams as input (AVC and MVC), and output a single interleaved stream with one frame from one stream followed by one frame from the other stream, and so on. The order of the images (AVC in odd and MVC in even frames or vice versa) must be controlled by a parameter. (Alternative outputs such as side by side or top & bottom combined frames can be considered if you wish, but since avisynth has the possibilities to build them from the interleaved stream, it's not necessary.)
As far as I know, the Intel example CLI command requires a single "combined" AVC+MVC video stream as input. I don't now if the library has the same limitation, but I suppose not, since the current MVCSource plugin can accept the 2 streams. (It can also accept a combined AVC+MVC single stream, but, unless it's easy to do, I don't think it's an useful feature.)
The current plugin can also decode and output the AVC stream only (but it is easy to decode it with other Source commands), or the MVC stream only (more useful, but not really necessary). It can also enable the hardware acceleration feature of the Intel library, but I suppose that that part can be added later, and I'm not sure it is really useful, as I don't know if the required hardware is frequent.
There is currently no audio support in MVCSource, and I don't think we need it.
Of course, it would be nice if the plugin could also accept a MPLS (or SSIF) file as input, but that requires probably some kind of demuxer, and I suppose it's a difficult and long work.
Another feature that is currently missing and that could be a very great improvement is some kind of seek support. Currently, all frames must be decoded sequentially. Not a big deal.