On Apr 18, 2006, at 5:27 PM, David E. Fox wrote:
On Tue, 18 Apr 2006 14:04:34 -0700 Andrew Sackville-West <andrew@farwestbilliards.com> wrote:I think the consensus was that some MUA's show it and some don't butthat mostly it was caused by pgp signing. That is a pgp-signed messageI surmised that after reading the thread - but ISTR being able to see the signature line fine on those that had GPG signed messages - maybe on another list (therefore using different list management software) -it would display the signature block and then any extra signature linesjust fine.I also noticed the signature missing on messages that are in part or inwhole HTML. In those cases there's a Text pane and a Attachment(s) pane, but no signature displays on either one.
And that is correct MUA behavior, even if it is not the intended _system_ behavior.
The unsubscribe signature/footer is either part of the message, or it is not. It is _not_ part of the message when it is sent to the list. The system _intent_ is that the footer be appended and distributed as part of the message.
If you look at the raw source of a problem message, regardless of MUA, you'll find that you're looking at a multipart mime message with
- 1. a text part, and - 2. one or more attachments (such as a GPG/PGP signature), and- 3. an epilogue containing the unsubscribe message (and, I'll wager it contains nothing else).
RFC2046 (see http://www.ietf.org/rfc/rfc2046.txt) is pretty clear on what constitutes the message, and what does not--check out the example at the bottom of page 20.
Given that the epilogue is in none of the message parts, it is _not_ part of the message. Since it is not part of the message, conforming MUAs do not display it as part of the message--what other behavior would you expect?
IOW, the list _system_ distributes multipart mime messages that are ill-formed for the intent of including the unsubscribe footer. Period. Any fault lies with whatever component appends the footer to the message--it has fumbled the execution of its intent. Period.
The MUA is irrelevant, though there are anecdotes of non-conforming MUAs which may mask the fault--if so, that is a double bug.
Fingering GPG/PGP as the cause demonstrates the result of sampling bias.