Ned Freed writes:
> I guess it would work - although it would probably trip loop detectors
> too soon in some cases - but I don't think an appproach that
> effectively calls for a message to lie about where it has been is a
> good idea.
AOL. There aren't any RFCs that say "Don't insert inaccurate Received
fields", but it still seems to have very bad karma ;)
> I think we need another mechanisms, probably header-based.
"Auto-Submitted: auto-generated" is defined in RFC 3834 section 5. 3834
orders autoresponders to be careful about incoming auto-submitted
messages. If similar language is added to the relevant sieve documents,
the result should be fine. At least in theory.
The main issue with using Auto-submitted: is that it would block all cases of
notifications sent in response to other notifications. Received: field counting
stops loops at some point but doesn't prevent a reasonable amount of
autoforwaring to occur before loop detection kicks in.
I just don't know if it is OK to block all notifications in response
to notifications. What do others think?
Ned
P.S. Full disclosure: Blocking notification responses to notifications would
eliminate a major layer 9 problem for our product team.