I'm a little bit of a man in the middle with this one so sorry if I sound
vague.
I've been maintaining an exim package for a small distro (Freesco) for
some time, currently it's using 3.36 (the change to 4.x will be soon) with
openSSL 0.9.6f. The "distro" is a Libc5 system with an old 2.0.38 kernel,
so I'm not sure if those are factors here or not.
Basically, a user of my package got in touch saying that exim wasn't
communicating with a communigate pro server correctly. Giving the error:
2002-10-09 19:07:48 blazej.runowski@??? T=remote_smtp defer (-19): Remote host pl86.szczecin.sdi.tpnet.pl [217.98.203.86] closed connection after end of data
2002-10-09 19:08:41 blazej.runowski@??? routing defer (-45): remote delivery suppressed
They contacted "Stalker" (someone to do with Communigate pro I imagine)
who apparently said:
> OK, we checked the code: there is only one situation when the SMTP
> module can generate this error - when the low-level routine returns 0
> read bytes. It can happen only if the sender has sent a TLS block with
> zero data. It's unlikely to happen with other algorithms, but with
> block-type algorithms like DES, it's easy to make a mistake on the
> sending end. As a result, it will send one full block with just
> "padding" symbols, and after decoding on the receiving end it will end
> up with the zero-size block. We can see if we can provide a workaround
> (i.e. ignore zero-size blocks), but you'll better notify the
> manufacturer of that "exim" program....
The closest I've found to any mention of something about this would be
http://www.exim.org/mailman/htdig/exim-users/Week-of-Mon-20020121/034418.html
"We've also had many hosts like charter.net with that stupid CommuniGate
Pro which defaults to advertising STARTTLS even if it's not configured to
do it, and when the said hosts fixed their stuff, the exim queue flushed
fine."
The host in question does advertise STARTTLS, but I'm not sure where the
problem actually lies here, given the response from "Stalker"
Would this actually be exim's problem or not?
DG
