So tell me, if one assumes the label approach (which I believe is present
in either proposal), why does that label have to be outside the content
itself? What's the difference, really, between a label and an HTML tag?
Because people get mad at MUA vendors when they see markup crud in their
mail messages. Same situation as QP encoding.
It's true that all proposals have some form of labeling. The MIME parameter
labeling is much more innocuous than a MIME type. Recall my experiment
where many MUA's failed miserably when presented with a text/paragraph
entity (even though failing so is a violation of the MIME spec).
Another way to look at this is that we have an installed base to deal with that
has certain features we can count on and certain features we cannot count on.
We can count on facilities to limit display of labels in headers. Headers
have been with us for a very long time and facilities exist in almost
every MUA out there to limit their display.
We cannot count on facilities to limit display of tags embedded in the actual
message text. It is true that MUAs haven't done their best to limit their use
of tags, but it is also true that the tolerance for tags in text is extremely
low.
Similarly, we cannot at this time deploy a new subtype of text without
problems. MIME implementations don't handle this sort of thing very well, the
MIME specification's requirements in this area notwithstanding.
Its been at least 10 years since these issues arose but it seems like nothing
has changed. These issues have also surfaced of late in regards to security
multiparts in the USEFOR WG. (I think a different response is indicated there,
BTW, but I don't want to get into that here.)
Ned