I agree that the particulars of how body parts are to be presented to a user
is up to the user agent. However, I wouldn't want to *recommend* that other
MIME implementations follow Andrew's convention of putting non-text body
parts on the same "line" as the preceeding text.
A .gif file, for example, is not a character. There's no information about
where the "baseline" is (i.e. where to place the bottom of the displayed
image relative to the text baseline), or how much room to leave on each side
of the displayed image when it appears next to text or other images.
Finally, while text size is defined in "points", that is, the size of the
characters when displayed, a .gif file has dimensions in terms of pixels.
Computer displays vary widely in resolution -- around here we have devices
ranging from ~ 50dpi to ~ 200dpi. So the ratio between text size and image
size might vary by as much as 4 to 1 on different displays, if there is no
way to indicate "this image should be displayed xxx points high."
My point is simply that additional information is required in order to
"properly" place a picture in-line with text. Andrew's hack is better than
not having this ability at all, but not general enough to be widely
interoperable.
Keith