ietf-822
[Top] [All Lists]

Re: General multipart structure (was MacMIME)

1993-03-18 05:41:26

On Thu, 18 Mar 1993 02:48:25 +0000, Harald Tveit Alvestrand 
<harald(_dot_)t(_dot_)alvestrand(_at_)delab(_dot_)sintef(_dot_)no> said:

        Harald> Why not define an extendible mechanism for Multipart:

        Harald> content-type=multipart/structured; control=first [or last]

        Harald> where we then have the semantics that:

        Harald> - A body part apart from the control has its normal meaning
        Harald> - The control body part has content that is about other parts.

        Harald> So, if you cannot parse the control, you can look at
        Harald> the other parts, but if you can parse the control, but
        Harald> not the other parts, you might as well junk the whole
        Harald> thing.

        ...

        Harald> (No, I did not suggest it for structuring Attachments.
        Harald> Others should get their fingers burned.....)

Harald, Dave, y'all:

I have been dwelling on this for a few days while sharpening the
Content-Disposition stuff, and I am beginning to becom convinced that
this way may be better, or at least complementary. I originally put
forth content-disposition because it provided a cleaner way to
represent dispositional info than mucking about with the message
structure, and because it gave a good location for other dispositional
info (filename, content caching hints, etc.) which just did not
belong anywhere else. These are all good things, and I still think
they are necessary.

But it is more important, I think, to have a unified view of multipart
files. This is true with PEM, with Mac files, and with structured
documents. I like your proposal, Harald, except that instead of
'first' or 'last' I would use the Content-ID of the organizer.
When handling this sort of thing Content-ID is not really optional.

_so_

How do people feel about this? Amd what is to be done with
Content-Disposition? I still see the need to indicate filename
and cahceing policy on a per-part basis. I'm not so sure about the
'inline' vs 'attachment' vs 'hidden' dispositional types given that
the same could be done with a (very simple) application subtype
for the organizer (what Harald terms the "control") defined in the
same document. Something like:

::::::::::
Content-Type: application/basic-organizer
Content-Transfer-Encoding: 7bit
Content-Disposition: inline;  
         filename="fubar";    
         cache=never;
Content-ID: <20026_1074_729552014_1(_at_)lorax>

Inline: <20026_1074_729552014_2(_at_)lorax>
Inline: <20026_1074_729552014_3(_at_)lorax>
Attachment: <20026_1074_729552014_4(_at_)lorax>
Attachment: <20026_1074_729552014_5(_at_)lorax>
Attachment: <20026_1074_729552014_6(_at_)lorax>
::::::::::


This way, my fingers won't be burned :)

Anyway, code is always better after it has been scrapped and
redesigned 3 or 4 times :( ++

-Rens
rens(_at_)cs(_dot_)columbia(_dot_)edu