As it turns out, the point with soon be moot w.r.t. Cyrus. I realized
last night that I never implemented ereject, and when I do, Cyrus will
reject at the LMTP level and therefore no DSN or MDN.
Yeah, I was going to mention that but forgot. This demonstrates a much more
fundamental problem with this whole idea of, "I want copies of all the messages
'the system' sends on my behalf." If a message is rejected due to use policy
encoded in sieve or wherever, it's always best to do that using a 5yz code over
protocol, which obviously provides no fcc capabilities.
This is also true if the reject occurs prior to delivery and the response is
sent over SMTP. So unless you say that the use of :fcc requires local
generation of the DSN - in the process making your system generate more
blowback spam that it would otherwise - you end up in a situation where the
user only receives copies of the subset of messages that happened to require
local DSN generation for some other reason.
And that's a pretty significant violation of the least astonishment principle.
I suppose you could do a hack where you generate your own version of DSN you
assume the SMTP client is going to create and stuff that in the user's folder.
But really, isn't this whole thing getting a little silly?
Ned
_______________________________________________
sieve mailing list
sieve(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/sieve