[BXXPwg] Mime headers

Marshall Rose mrose+mtr.netnews@dbc.mtview.ca.us
Wed, 19 Jul 2000 22:24:23 -0700


> At 10:20 AM 7/17/00 -0700, Marshall Rose wrote:
> >i don't think so. i think you can send mime headers only on the first
> >segment of a message. it doesn't make sense to have them anywhere else.
we
> >should probably add a consistency check rule for that.
>
> Eh?  Unless I'm missing something, the "consistency rule" is a valid MIME
> object.  It seems to me quite legitimate (though probably unnecessary) to
> have MIME headers overflow the first segment.
>
> I would expect the concatenation of segment data to be treated as a single
> MIME object, and analyzed accordingly.  (This raises the interesting point
> that the default text/xml MIME type should not be assumed until the first
> blank line after the MIME headers has been seen.

there's a bit of context i think you're missing here.

the term "consistency rules" is a shorthand used on the mailing list that
refers to the set of rules starting on page 10 of Section 2.2.1.1 that
explain how to look at a frame header and decide if you're playing the right
game. right now there are nine rules...

the thing to keep in mind is that the MIME headers are not part of the
payload, they're part of the frame header. the payload is the body. there
are pluses and minuses to this design point. (the reason it's a design point
is that the header is designed to put all the useful stuff up front where
you can find "fairly easily"...for example, we don't use a Content-Length:
header to say how big the payload is, we just put the number of octets in
the parameters carried by the header).

the plus is that it forces the sender to put all the MIME headers in the
first segment so the other side can look at it and decide what to do.

the minus is that the headers aren't included in things like octet counts,
etc.

this may be an over-optimization (like the piggybacking thing we talked
about yesterday). i'm not sure. but, i really like the fact that a receiver
can get the first segment of a message in and know exactly what the sender
is trying to give it.

so, to return to Sarveshwar Rao Duddu's original point: the MIME headers --
as the spec is currently written -- should be discussed in "the consistency
rules"...

/mtr