Setting aside the discussion of whether digests are "good, bad, or
ugly", I am very interested in whether or not there is any consensus as
to what is an ideal format for a digest given that it will exist.
I want to use multipart/digest. Here is what I have understood from the
discussion thus far.
From: (see point 1 below)
To: SUBMISSION_ADDRESS_FOR_DIGEST
Subject: Digest #??
Content-type: multipart/digest; boundary="line1"
--line1
Subject: Table of Contents (see point 2 below)
LIST_OF_SUBJECTS (see point 3 below)
--line1
HEADERS_OF_DIGEST_MESSAGE #1 (see point 4 below)
BODY_OF_DIGEST_MESSAGE #1
--line1
HEADERS_OF_DIGEST_MESSAGE #2
BODY_OF_DIGEST_MESSAGE #2
--line1--
Point 1:
In my digests I set the From: message header to be the errors
address for the list, making the default assumption that only error
messages are "returned or replied" and choosing to filter out
messages intended for the list. My statistics are excellent in this
regard. It's also been suggested you can make it the digest
submission address and instead filter out messages that do not
belong on the list. With the right filter either setting should be
equally good.
Point 2:
I incorrectly made the point that an embedded message/rfc822 needs
to have a From:, Subject:, and Date:. I checked again, carefully,
and that is not what rfc2046 says. It very carefully says that an
outermost message/rfc822 content-type needs must have all of a
message From:, Date:, and a destination address header present. In
contrast, the encapsulated headers of an embedded message/rfc822
must have at least one of a message From:, Subject:, or Date:
header.
So, in the interests of not having a multipart in a multipart and
thus keeping the table of contents with the digest, the Subject:
header seems like a good candidate as the minimum header.
Point 3:
In my digests the list of subjects is exactly that, just a list of
the subject headers from the enclosed messages, one per line.
However, I don't see any reason why a multipart/digest could not
have exactly one part in it which was the list of subject header
values with each one linked to the actual message, as suggested
elsewhere on this mailing list. In fact, I like this idea.
Point 4:
Traditionally, digests have been formatted such that an explicit
list of headers has been included. In my digests I include all
headers except Received: lines, believing those to be more space
consuming than content rich as far as a digest is concerned. Thus,
an email client or archive technology has available to them all the
information they need for replies or threading or whatever other
functionality they wish to offer.
Comments, questions, or suggestions?
Jim
