First I would like to thank those who took time to write long articles in
anwer to my questions. I am glad I raised the issue : while I do understand
that the whole affair is indeed technically coherent, I do believe that we
will see incorrect implementations if the wording of the specification is
not further refined.
Please remember that the whole world is not made of Unix machines. There
are some systems where the MUA and MTA and completely separate, and where
mail items are transmitted asynchronously between them. In some systems I
know and have experience with, the current practice is that the transmission
between MUA and MTA is done using general purpose channels available in
the OS, and the transmission of the text of the message (header+content) is
actually done in the normal internal format of the machine. Only when actual
transmission is done by the SMTP component of the MTA are Internet text
convention applied/removed. In such a setup, the MUA never sees 'Internet
standard text format'.
With the advent of MIME, and the necessity to encode some texts to allow
arbitrary character sets and/or long lines, the MUA will have to deal
with a transformation of text between 'internal' and 'Internet standard'
in some cases, but no necessarily always. For example, if the MTA is not
to be modified (the primary goal of MIME...), it follows that the MUA
will have to convert text to 'Internet format' when it must be encoded
only ; and, of course, the result of the encoding operation will be
produced and transmitted to the MTA in 'internal' format, as will be
non-encoded texts.
All this if of course doable. But in the kind of setup I described above,
it certainly is a completly new adventure for the MUA. So it seem dangerous
(at least to my limited intelligence) to let this happen only by
understatement (the kind : Internet standard text has always been this and
that and MIME did not change anything...). I'll propose some editorial
sugar in the very near future. /AF