Arnt Gulbrandsen <arnt(_at_)gulbrandsen(_dot_)priv(_dot_)no> writes:
Simon Josefsson writes:
Arnt Gulbrandsen <arnt(_at_)gulbrandsen(_dot_)priv(_dot_)no> writes:
Mail delivery notification needs some sort of authentication, and
it needs knowledge of when which mailboxes receive mail. Sieve has
the latter.
That doesn't imply the managesieve server has that knowledge. I
think that for all it is worth, the managesieve could be located on
an entirely different machine and not at all connected to the
mail-delivery process. It might just be responsible for copying the
sieve script, and switch between the active scripts, on the mail
server.
The managesieve server certainly knows which mailboxes are delivery
candidates. To know that, one has to parse the Sieve scripts.
I thought the managesieve server only parsed the script for syntax
errors. I.e., managesieve is FTP with a built-in managesieve parser.
Can you point to anything in the specification that say the
managesieve server know about whether delivery candidates exists or
not? I haven't seen a server reject scripts if I make a typo for the
mailboxes.
So, are two TCP protocols for talking to Sieve software desirable?
I say that's one too much, which leaves me with the suggestion:
Managesieve should offer mail delivery nofication. Ridiculous.
I believe Managesieve is useful, and that a mail delivery protocol
is also useful. Whether the mail delivery protocol should be
sieve-specific isn't evident to me. Overloading managesieve for this
purpose seem more complicated than inventing a new protocol.
ANOTHER protocol with exactly the same authentication/tls verbiage?
Yes, or re-use a perhaps more suitable protocol, such as IMAP, SIP,
XMPP or Zephyr. Given existing TLS/SASL libraries, supporting
authentication is fortunately rather easy for new protocols these
days.
Regards,
Simon