Having looked at the cited part of the RFC (but not tried to analyze it in detail), I think you are correct.  I've also glanced at your PR, and I think your approach is correct in broad outline, but I haven't looked at the details.  For full message/global support, however, it will also be necessary to look at the output side: given a message/global part, a transfer encoding should be applied when serializing with cte_type=7bit.  Support for message/global should also be added to the contentmanager.

I won't have an objection if this is accepted with only the feedparser support, but I would recommend that the remaining pieces of support for message/global be added before the feature is released.