Wait. I think some people have missed Vincent's point, and have
gotten too wrapped up in the particular example.
The example of nroff was *not* proposing nroff as the "net standard
formatting language". It was only trying to be an example of an older
program which does not internally identify the character set that it
uses; therefore the character set needs to be represented on the outside
of the body part.
There have been two proposed solutions: concatenate the character
set identifier along with the type name, or represent the character
set in a separate header field of its own.
These solutions are clearly equivalent in the information transmission
sense -- anything that one can represent the other can too. The choice
between them will be made because of differences in style and religion
rather than because "something couldn't be done" one way or the other.
With that said, I'ld like to state my opinion:
I think that the "character set encoding" information should
be split to a different field. Why? I have a model where there
are pieces of information that the "mail system" wants to understand,
independent of the data type -- things like encoding method(s),
compression method(s), character encodings, etc. This is all
information on the body part "envelope"; things that are carried
around in addition to the actual data.
Just today we've already heard two different positional combinations
of the type field: type/character_set and type/compression. One
can easily believe that new things will need to be standardized.
The easiest way to do this is to give these attributes their own
header field in the body part, rather than trying to positionally
represent them in the type field.
About Nathaniel's point with respect to nroff being "old technology"
that doesn't internally identify a character set (whereas things like
postscript, Andrew, etc, internally represent the character set) --
like it or not these things abound in today's world, and we should
make sure that we pass enough information to properly display the
document on the user's screen.
Neil