Personally I am skeptical if suggesting UTF-8 for the body data is a good idea, but I can go along with it, since other people want it. But I do strongly question whether it is right to suggest UTF-8 for header fields. The RFC <https://tools.ietf.org/html/rfc7230#page-26> only mentions ASCII and Latin-1.
Newer protocols based on HTTP (RTSP comes to mind) do specify UTF-8 for the header, but that is probably out of scope of both the HTTP module and beginner-targetted errors.
If I were redoing this patch, I would drop all the changes except at the body.encode() line in Emil’s original post.