Folks,
While reviewing the mailing list archive I've stumbled across an old
message from Nigel Swinson which I don't think received adequate
considerations. Nigel wrote in August 2007:
6. Action Enclose
[...]
The text implies that enclose happens immediately, and therefore subsequent
actions affecting the message header/content apply to the enclosing message,
but then it goes on to say if there are multiple enclose actions, then it's
only the last one that will win. This means if I have an enclose action,
I'll have to remember that the outer body part is the new "enclosing" body
part, and should I get another enclose action then I should "replace" that
enclosing body part rather than enclose the message yet again such that it
has two layers of enclosing.
This sounds like a real hassle to implement for questionable utility, I'd
rather two enclose actions just add two layers of enclosing mime structure
and body text body parts be present. Just like they would if the message
goes through two separate Sieve scripts, both of which add an enclose.
Speaking as an implementor I would like to second this. Do we need this
extra complexity? If an implementation doesn't want to enclose a
bodypart twice, it can either use Sieve variables, or construct a
different Sieve script.
Either that or define that the enclose action should happen at the end of
the script and therefore doesn't affect the mime structure during the
existing processing (that might even be inside a for_every_part construct).
I'd actually suggest we leave it to the implementation like we do for
fileinto. The implementation can choose to have the enclose happen
immediately, or at the end.